]>
Commit | Line | Data |
---|---|---|
749cf76c CD |
1 | /* |
2 | * Copyright (C) 2012 - Virtual Open Systems and Columbia University | |
3 | * Author: Christoffer Dall <c.dall@virtualopensystems.com> | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify | |
6 | * it under the terms of the GNU General Public License, version 2, as | |
7 | * published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope that it will be useful, | |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
13 | * | |
14 | * You should have received a copy of the GNU General Public License | |
15 | * along with this program; if not, write to the Free Software | |
16 | * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. | |
17 | */ | |
18 | ||
85acda3b | 19 | #include <linux/bug.h> |
1fcf7ce0 | 20 | #include <linux/cpu_pm.h> |
749cf76c CD |
21 | #include <linux/errno.h> |
22 | #include <linux/err.h> | |
23 | #include <linux/kvm_host.h> | |
1085fdc6 | 24 | #include <linux/list.h> |
749cf76c CD |
25 | #include <linux/module.h> |
26 | #include <linux/vmalloc.h> | |
27 | #include <linux/fs.h> | |
28 | #include <linux/mman.h> | |
29 | #include <linux/sched.h> | |
86ce8535 | 30 | #include <linux/kvm.h> |
2412405b EA |
31 | #include <linux/kvm_irqfd.h> |
32 | #include <linux/irqbypass.h> | |
de737089 | 33 | #include <linux/sched/stat.h> |
749cf76c | 34 | #include <trace/events/kvm.h> |
b02386eb | 35 | #include <kvm/arm_pmu.h> |
1a2fb94e | 36 | #include <kvm/arm_psci.h> |
749cf76c CD |
37 | |
38 | #define CREATE_TRACE_POINTS | |
39 | #include "trace.h" | |
40 | ||
7c0f6ba6 | 41 | #include <linux/uaccess.h> |
749cf76c CD |
42 | #include <asm/ptrace.h> |
43 | #include <asm/mman.h> | |
342cd0ab | 44 | #include <asm/tlbflush.h> |
5b3e5e5b | 45 | #include <asm/cacheflush.h> |
85acda3b | 46 | #include <asm/cpufeature.h> |
342cd0ab CD |
47 | #include <asm/virt.h> |
48 | #include <asm/kvm_arm.h> | |
49 | #include <asm/kvm_asm.h> | |
50 | #include <asm/kvm_mmu.h> | |
f7ed45be | 51 | #include <asm/kvm_emulate.h> |
5b3e5e5b | 52 | #include <asm/kvm_coproc.h> |
910917bb | 53 | #include <asm/sections.h> |
749cf76c CD |
54 | |
55 | #ifdef REQUIRES_VIRT | |
56 | __asm__(".arch_extension virt"); | |
57 | #endif | |
58 | ||
36989e7f | 59 | DEFINE_PER_CPU(kvm_cpu_context_t, kvm_host_cpu_state); |
342cd0ab | 60 | static DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page); |
342cd0ab | 61 | |
1638a12d MZ |
62 | /* Per-CPU variable containing the currently running vcpu. */ |
63 | static DEFINE_PER_CPU(struct kvm_vcpu *, kvm_arm_running_vcpu); | |
64 | ||
f7ed45be CD |
65 | /* The VMID used in the VTTBR */ |
66 | static atomic64_t kvm_vmid_gen = ATOMIC64_INIT(1); | |
20475f78 VM |
67 | static u32 kvm_next_vmid; |
68 | static unsigned int kvm_vmid_bits __read_mostly; | |
f0cf47d9 | 69 | static DEFINE_RWLOCK(kvm_vmid_lock); |
342cd0ab | 70 | |
c7da6fa4 PF |
71 | static bool vgic_present; |
72 | ||
67f69197 AT |
73 | static DEFINE_PER_CPU(unsigned char, kvm_arm_hardware_enabled); |
74 | ||
1638a12d MZ |
75 | static void kvm_arm_set_running_vcpu(struct kvm_vcpu *vcpu) |
76 | { | |
1436c1aa | 77 | __this_cpu_write(kvm_arm_running_vcpu, vcpu); |
1638a12d MZ |
78 | } |
79 | ||
61bbe380 CD |
80 | DEFINE_STATIC_KEY_FALSE(userspace_irqchip_in_use); |
81 | ||
1638a12d MZ |
82 | /** |
83 | * kvm_arm_get_running_vcpu - get the vcpu running on the current CPU. | |
84 | * Must be called from non-preemptible context | |
85 | */ | |
86 | struct kvm_vcpu *kvm_arm_get_running_vcpu(void) | |
87 | { | |
1436c1aa | 88 | return __this_cpu_read(kvm_arm_running_vcpu); |
1638a12d MZ |
89 | } |
90 | ||
91 | /** | |
92 | * kvm_arm_get_running_vcpus - get the per-CPU array of currently running vcpus. | |
93 | */ | |
4000be42 | 94 | struct kvm_vcpu * __percpu *kvm_get_running_vcpus(void) |
1638a12d MZ |
95 | { |
96 | return &kvm_arm_running_vcpu; | |
97 | } | |
98 | ||
749cf76c CD |
99 | int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu) |
100 | { | |
101 | return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE; | |
102 | } | |
103 | ||
749cf76c CD |
104 | int kvm_arch_hardware_setup(void) |
105 | { | |
106 | return 0; | |
107 | } | |
108 | ||
749cf76c CD |
109 | void kvm_arch_check_processor_compat(void *rtn) |
110 | { | |
111 | *(int *)rtn = 0; | |
112 | } | |
113 | ||
749cf76c | 114 | |
d5d8184d CD |
115 | /** |
116 | * kvm_arch_init_vm - initializes a VM data structure | |
117 | * @kvm: pointer to the KVM struct | |
118 | */ | |
749cf76c CD |
119 | int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) |
120 | { | |
94d0e598 | 121 | int ret, cpu; |
d5d8184d | 122 | |
5b6c6742 SP |
123 | ret = kvm_arm_config_vm(kvm, type); |
124 | if (ret) | |
125 | return ret; | |
749cf76c | 126 | |
94d0e598 MZ |
127 | kvm->arch.last_vcpu_ran = alloc_percpu(typeof(*kvm->arch.last_vcpu_ran)); |
128 | if (!kvm->arch.last_vcpu_ran) | |
129 | return -ENOMEM; | |
130 | ||
131 | for_each_possible_cpu(cpu) | |
132 | *per_cpu_ptr(kvm->arch.last_vcpu_ran, cpu) = -1; | |
133 | ||
d5d8184d CD |
134 | ret = kvm_alloc_stage2_pgd(kvm); |
135 | if (ret) | |
136 | goto out_fail_alloc; | |
137 | ||
c8dddecd | 138 | ret = create_hyp_mappings(kvm, kvm + 1, PAGE_HYP); |
d5d8184d CD |
139 | if (ret) |
140 | goto out_free_stage2_pgd; | |
141 | ||
6c3d63c9 | 142 | kvm_vgic_early_init(kvm); |
a1a64387 | 143 | |
d5d8184d CD |
144 | /* Mark the initial VMID generation invalid */ |
145 | kvm->arch.vmid_gen = 0; | |
146 | ||
3caa2d8c | 147 | /* The maximum number of VCPUs is limited by the host's GIC model */ |
c7da6fa4 PF |
148 | kvm->arch.max_vcpus = vgic_present ? |
149 | kvm_vgic_get_max_vcpus() : KVM_MAX_VCPUS; | |
3caa2d8c | 150 | |
d5d8184d CD |
151 | return ret; |
152 | out_free_stage2_pgd: | |
153 | kvm_free_stage2_pgd(kvm); | |
154 | out_fail_alloc: | |
94d0e598 MZ |
155 | free_percpu(kvm->arch.last_vcpu_ran); |
156 | kvm->arch.last_vcpu_ran = NULL; | |
d5d8184d | 157 | return ret; |
749cf76c CD |
158 | } |
159 | ||
235539b4 LC |
160 | bool kvm_arch_has_vcpu_debugfs(void) |
161 | { | |
162 | return false; | |
163 | } | |
164 | ||
165 | int kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu) | |
166 | { | |
167 | return 0; | |
168 | } | |
169 | ||
1499fa80 | 170 | vm_fault_t kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf) |
749cf76c CD |
171 | { |
172 | return VM_FAULT_SIGBUS; | |
173 | } | |
174 | ||
749cf76c | 175 | |
d5d8184d CD |
176 | /** |
177 | * kvm_arch_destroy_vm - destroy the VM data structure | |
178 | * @kvm: pointer to the KVM struct | |
179 | */ | |
749cf76c CD |
180 | void kvm_arch_destroy_vm(struct kvm *kvm) |
181 | { | |
182 | int i; | |
183 | ||
b2c9a85d MZ |
184 | kvm_vgic_destroy(kvm); |
185 | ||
94d0e598 MZ |
186 | free_percpu(kvm->arch.last_vcpu_ran); |
187 | kvm->arch.last_vcpu_ran = NULL; | |
188 | ||
749cf76c CD |
189 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
190 | if (kvm->vcpus[i]) { | |
191 | kvm_arch_vcpu_free(kvm->vcpus[i]); | |
192 | kvm->vcpus[i] = NULL; | |
193 | } | |
194 | } | |
6b2ad81b | 195 | atomic_set(&kvm->online_vcpus, 0); |
749cf76c CD |
196 | } |
197 | ||
784aa3d7 | 198 | int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext) |
749cf76c CD |
199 | { |
200 | int r; | |
201 | switch (ext) { | |
1a89dd91 | 202 | case KVM_CAP_IRQCHIP: |
c7da6fa4 PF |
203 | r = vgic_present; |
204 | break; | |
d44758c0 | 205 | case KVM_CAP_IOEVENTFD: |
7330672b | 206 | case KVM_CAP_DEVICE_CTRL: |
749cf76c CD |
207 | case KVM_CAP_USER_MEMORY: |
208 | case KVM_CAP_SYNC_MMU: | |
209 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: | |
210 | case KVM_CAP_ONE_REG: | |
aa024c2f | 211 | case KVM_CAP_ARM_PSCI: |
4447a208 | 212 | case KVM_CAP_ARM_PSCI_0_2: |
98047888 | 213 | case KVM_CAP_READONLY_MEM: |
ecccf0cc | 214 | case KVM_CAP_MP_STATE: |
460df4c1 | 215 | case KVM_CAP_IMMEDIATE_EXIT: |
749cf76c CD |
216 | r = 1; |
217 | break; | |
3401d546 CD |
218 | case KVM_CAP_ARM_SET_DEVICE_ADDR: |
219 | r = 1; | |
ca46e10f | 220 | break; |
749cf76c CD |
221 | case KVM_CAP_NR_VCPUS: |
222 | r = num_online_cpus(); | |
223 | break; | |
224 | case KVM_CAP_MAX_VCPUS: | |
225 | r = KVM_MAX_VCPUS; | |
226 | break; | |
7af4df85 LC |
227 | case KVM_CAP_NR_MEMSLOTS: |
228 | r = KVM_USER_MEM_SLOTS; | |
229 | break; | |
2988509d VM |
230 | case KVM_CAP_MSI_DEVID: |
231 | if (!kvm) | |
232 | r = -EINVAL; | |
233 | else | |
234 | r = kvm->arch.vgic.msis_require_devid; | |
235 | break; | |
f7214e60 CD |
236 | case KVM_CAP_ARM_USER_IRQ: |
237 | /* | |
238 | * 1: EL1_VTIMER, EL1_PTIMER, and PMU. | |
239 | * (bump this number if adding more devices) | |
240 | */ | |
241 | r = 1; | |
242 | break; | |
749cf76c | 243 | default: |
b46f01ce | 244 | r = kvm_arch_dev_ioctl_check_extension(kvm, ext); |
749cf76c CD |
245 | break; |
246 | } | |
247 | return r; | |
248 | } | |
249 | ||
250 | long kvm_arch_dev_ioctl(struct file *filp, | |
251 | unsigned int ioctl, unsigned long arg) | |
252 | { | |
253 | return -EINVAL; | |
254 | } | |
255 | ||
d1e5b0e9 MO |
256 | struct kvm *kvm_arch_alloc_vm(void) |
257 | { | |
258 | if (!has_vhe()) | |
259 | return kzalloc(sizeof(struct kvm), GFP_KERNEL); | |
260 | ||
261 | return vzalloc(sizeof(struct kvm)); | |
262 | } | |
263 | ||
264 | void kvm_arch_free_vm(struct kvm *kvm) | |
265 | { | |
266 | if (!has_vhe()) | |
267 | kfree(kvm); | |
268 | else | |
269 | vfree(kvm); | |
270 | } | |
749cf76c CD |
271 | |
272 | struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id) | |
273 | { | |
274 | int err; | |
275 | struct kvm_vcpu *vcpu; | |
276 | ||
716139df CD |
277 | if (irqchip_in_kernel(kvm) && vgic_initialized(kvm)) { |
278 | err = -EBUSY; | |
279 | goto out; | |
280 | } | |
281 | ||
3caa2d8c AP |
282 | if (id >= kvm->arch.max_vcpus) { |
283 | err = -EINVAL; | |
284 | goto out; | |
285 | } | |
286 | ||
749cf76c CD |
287 | vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL); |
288 | if (!vcpu) { | |
289 | err = -ENOMEM; | |
290 | goto out; | |
291 | } | |
292 | ||
293 | err = kvm_vcpu_init(vcpu, kvm, id); | |
294 | if (err) | |
295 | goto free_vcpu; | |
296 | ||
c8dddecd | 297 | err = create_hyp_mappings(vcpu, vcpu + 1, PAGE_HYP); |
d5d8184d CD |
298 | if (err) |
299 | goto vcpu_uninit; | |
300 | ||
749cf76c | 301 | return vcpu; |
d5d8184d CD |
302 | vcpu_uninit: |
303 | kvm_vcpu_uninit(vcpu); | |
749cf76c CD |
304 | free_vcpu: |
305 | kmem_cache_free(kvm_vcpu_cache, vcpu); | |
306 | out: | |
307 | return ERR_PTR(err); | |
308 | } | |
309 | ||
31928aa5 | 310 | void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu) |
749cf76c | 311 | { |
749cf76c CD |
312 | } |
313 | ||
314 | void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu) | |
315 | { | |
f1d7231c CD |
316 | if (vcpu->arch.has_run_once && unlikely(!irqchip_in_kernel(vcpu->kvm))) |
317 | static_branch_dec(&userspace_irqchip_in_use); | |
318 | ||
d5d8184d | 319 | kvm_mmu_free_memory_caches(vcpu); |
967f8427 | 320 | kvm_timer_vcpu_terminate(vcpu); |
5f0a714a | 321 | kvm_pmu_vcpu_destroy(vcpu); |
591d215a | 322 | kvm_vcpu_uninit(vcpu); |
d5d8184d | 323 | kmem_cache_free(kvm_vcpu_cache, vcpu); |
749cf76c CD |
324 | } |
325 | ||
326 | void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu) | |
327 | { | |
328 | kvm_arch_vcpu_free(vcpu); | |
329 | } | |
330 | ||
331 | int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu) | |
332 | { | |
1c88ab7e | 333 | return kvm_timer_is_pending(vcpu); |
749cf76c CD |
334 | } |
335 | ||
d35268da CD |
336 | void kvm_arch_vcpu_blocking(struct kvm_vcpu *vcpu) |
337 | { | |
338 | kvm_timer_schedule(vcpu); | |
df9ba959 | 339 | kvm_vgic_v4_enable_doorbell(vcpu); |
d35268da CD |
340 | } |
341 | ||
342 | void kvm_arch_vcpu_unblocking(struct kvm_vcpu *vcpu) | |
343 | { | |
344 | kvm_timer_unschedule(vcpu); | |
df9ba959 | 345 | kvm_vgic_v4_disable_doorbell(vcpu); |
d35268da CD |
346 | } |
347 | ||
749cf76c CD |
348 | int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu) |
349 | { | |
f7ed45be CD |
350 | /* Force users to call KVM_ARM_VCPU_INIT */ |
351 | vcpu->arch.target = -1; | |
f7fa034d | 352 | bitmap_zero(vcpu->arch.features, KVM_VCPU_MAX_FEATURES); |
1a89dd91 | 353 | |
967f8427 MZ |
354 | /* Set up the timer */ |
355 | kvm_timer_vcpu_init(vcpu); | |
356 | ||
84e690bf AB |
357 | kvm_arm_reset_debug_ptr(vcpu); |
358 | ||
1aab6f46 | 359 | return kvm_vgic_vcpu_init(vcpu); |
749cf76c CD |
360 | } |
361 | ||
749cf76c CD |
362 | void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu) |
363 | { | |
94d0e598 MZ |
364 | int *last_ran; |
365 | ||
366 | last_ran = this_cpu_ptr(vcpu->kvm->arch.last_vcpu_ran); | |
367 | ||
368 | /* | |
369 | * We might get preempted before the vCPU actually runs, but | |
370 | * over-invalidation doesn't affect correctness. | |
371 | */ | |
372 | if (*last_ran != vcpu->vcpu_id) { | |
373 | kvm_call_hyp(__kvm_tlb_flush_local_vmid, vcpu); | |
374 | *last_ran = vcpu->vcpu_id; | |
375 | } | |
376 | ||
86ce8535 | 377 | vcpu->cpu = cpu; |
36989e7f | 378 | vcpu->arch.host_cpu_context = this_cpu_ptr(&kvm_host_cpu_state); |
5b3e5e5b | 379 | |
1638a12d | 380 | kvm_arm_set_running_vcpu(vcpu); |
328e5664 | 381 | kvm_vgic_load(vcpu); |
b103cc3f | 382 | kvm_timer_vcpu_load(vcpu); |
bc192cee | 383 | kvm_vcpu_load_sysregs(vcpu); |
e6b673b7 | 384 | kvm_arch_vcpu_load_fp(vcpu); |
de737089 MZ |
385 | |
386 | if (single_task_running()) | |
387 | vcpu_clear_wfe_traps(vcpu); | |
388 | else | |
389 | vcpu_set_wfe_traps(vcpu); | |
749cf76c CD |
390 | } |
391 | ||
392 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) | |
393 | { | |
e6b673b7 | 394 | kvm_arch_vcpu_put_fp(vcpu); |
bc192cee | 395 | kvm_vcpu_put_sysregs(vcpu); |
b103cc3f | 396 | kvm_timer_vcpu_put(vcpu); |
328e5664 CD |
397 | kvm_vgic_put(vcpu); |
398 | ||
e9b152cb CD |
399 | vcpu->cpu = -1; |
400 | ||
1638a12d | 401 | kvm_arm_set_running_vcpu(NULL); |
749cf76c CD |
402 | } |
403 | ||
424c989b AJ |
404 | static void vcpu_power_off(struct kvm_vcpu *vcpu) |
405 | { | |
406 | vcpu->arch.power_off = true; | |
7b244e2b | 407 | kvm_make_request(KVM_REQ_SLEEP, vcpu); |
424c989b AJ |
408 | kvm_vcpu_kick(vcpu); |
409 | } | |
410 | ||
749cf76c CD |
411 | int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, |
412 | struct kvm_mp_state *mp_state) | |
413 | { | |
3781528e | 414 | if (vcpu->arch.power_off) |
ecccf0cc AB |
415 | mp_state->mp_state = KVM_MP_STATE_STOPPED; |
416 | else | |
417 | mp_state->mp_state = KVM_MP_STATE_RUNNABLE; | |
418 | ||
419 | return 0; | |
749cf76c CD |
420 | } |
421 | ||
422 | int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, | |
423 | struct kvm_mp_state *mp_state) | |
424 | { | |
e83dff5e CD |
425 | int ret = 0; |
426 | ||
ecccf0cc AB |
427 | switch (mp_state->mp_state) { |
428 | case KVM_MP_STATE_RUNNABLE: | |
3781528e | 429 | vcpu->arch.power_off = false; |
ecccf0cc AB |
430 | break; |
431 | case KVM_MP_STATE_STOPPED: | |
424c989b | 432 | vcpu_power_off(vcpu); |
ecccf0cc AB |
433 | break; |
434 | default: | |
e83dff5e | 435 | ret = -EINVAL; |
ecccf0cc AB |
436 | } |
437 | ||
e83dff5e | 438 | return ret; |
749cf76c CD |
439 | } |
440 | ||
5b3e5e5b CD |
441 | /** |
442 | * kvm_arch_vcpu_runnable - determine if the vcpu can be scheduled | |
443 | * @v: The VCPU pointer | |
444 | * | |
445 | * If the guest CPU is not waiting for interrupts or an interrupt line is | |
446 | * asserted, the CPU is by definition runnable. | |
447 | */ | |
749cf76c CD |
448 | int kvm_arch_vcpu_runnable(struct kvm_vcpu *v) |
449 | { | |
3df59d8d CD |
450 | bool irq_lines = *vcpu_hcr(v) & (HCR_VI | HCR_VF); |
451 | return ((irq_lines || kvm_vgic_vcpu_pending_irq(v)) | |
3b92830a | 452 | && !v->arch.power_off && !v->arch.pause); |
749cf76c CD |
453 | } |
454 | ||
199b5763 LM |
455 | bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu) |
456 | { | |
f01fbd2f | 457 | return vcpu_mode_priv(vcpu); |
199b5763 LM |
458 | } |
459 | ||
f7ed45be CD |
460 | /* Just ensure a guest exit from a particular CPU */ |
461 | static void exit_vm_noop(void *info) | |
462 | { | |
463 | } | |
464 | ||
465 | void force_vm_exit(const cpumask_t *mask) | |
466 | { | |
898f949f | 467 | preempt_disable(); |
f7ed45be | 468 | smp_call_function_many(mask, exit_vm_noop, NULL, true); |
898f949f | 469 | preempt_enable(); |
f7ed45be CD |
470 | } |
471 | ||
472 | /** | |
473 | * need_new_vmid_gen - check that the VMID is still valid | |
6a727b0b | 474 | * @kvm: The VM's VMID to check |
f7ed45be CD |
475 | * |
476 | * return true if there is a new generation of VMIDs being used | |
477 | * | |
478 | * The hardware supports only 256 values with the value zero reserved for the | |
479 | * host, so we check if an assigned value belongs to a previous generation, | |
480 | * which which requires us to assign a new value. If we're the first to use a | |
481 | * VMID for the new generation, we must flush necessary caches and TLBs on all | |
482 | * CPUs. | |
483 | */ | |
484 | static bool need_new_vmid_gen(struct kvm *kvm) | |
485 | { | |
486 | return unlikely(kvm->arch.vmid_gen != atomic64_read(&kvm_vmid_gen)); | |
487 | } | |
488 | ||
489 | /** | |
490 | * update_vttbr - Update the VTTBR with a valid VMID before the guest runs | |
491 | * @kvm The guest that we are about to run | |
492 | * | |
493 | * Called from kvm_arch_vcpu_ioctl_run before entering the guest to ensure the | |
494 | * VM has a valid VMID, otherwise assigns a new one and flushes corresponding | |
495 | * caches and TLBs. | |
496 | */ | |
497 | static void update_vttbr(struct kvm *kvm) | |
498 | { | |
499 | phys_addr_t pgd_phys; | |
500 | u64 vmid; | |
f0cf47d9 | 501 | bool new_gen; |
f7ed45be | 502 | |
f0cf47d9 MZ |
503 | read_lock(&kvm_vmid_lock); |
504 | new_gen = need_new_vmid_gen(kvm); | |
505 | read_unlock(&kvm_vmid_lock); | |
506 | ||
507 | if (!new_gen) | |
f7ed45be CD |
508 | return; |
509 | ||
f0cf47d9 | 510 | write_lock(&kvm_vmid_lock); |
f7ed45be CD |
511 | |
512 | /* | |
513 | * We need to re-check the vmid_gen here to ensure that if another vcpu | |
514 | * already allocated a valid vmid for this vm, then this vcpu should | |
515 | * use the same vmid. | |
516 | */ | |
517 | if (!need_new_vmid_gen(kvm)) { | |
f0cf47d9 | 518 | write_unlock(&kvm_vmid_lock); |
f7ed45be CD |
519 | return; |
520 | } | |
521 | ||
522 | /* First user of a new VMID generation? */ | |
523 | if (unlikely(kvm_next_vmid == 0)) { | |
524 | atomic64_inc(&kvm_vmid_gen); | |
525 | kvm_next_vmid = 1; | |
526 | ||
527 | /* | |
528 | * On SMP we know no other CPUs can use this CPU's or each | |
529 | * other's VMID after force_vm_exit returns since the | |
530 | * kvm_vmid_lock blocks them from reentry to the guest. | |
531 | */ | |
532 | force_vm_exit(cpu_all_mask); | |
533 | /* | |
534 | * Now broadcast TLB + ICACHE invalidation over the inner | |
535 | * shareable domain to make sure all data structures are | |
536 | * clean. | |
537 | */ | |
538 | kvm_call_hyp(__kvm_flush_vm_context); | |
539 | } | |
540 | ||
541 | kvm->arch.vmid_gen = atomic64_read(&kvm_vmid_gen); | |
542 | kvm->arch.vmid = kvm_next_vmid; | |
543 | kvm_next_vmid++; | |
20475f78 | 544 | kvm_next_vmid &= (1 << kvm_vmid_bits) - 1; |
f7ed45be CD |
545 | |
546 | /* update vttbr to be used with the new vmid */ | |
9163ee23 | 547 | pgd_phys = virt_to_phys(kvm->arch.pgd); |
e55cac5b | 548 | BUG_ON(pgd_phys & ~kvm_vttbr_baddr_mask(kvm)); |
20475f78 | 549 | vmid = ((u64)(kvm->arch.vmid) << VTTBR_VMID_SHIFT) & VTTBR_VMID_MASK(kvm_vmid_bits); |
529c4b05 | 550 | kvm->arch.vttbr = kvm_phys_to_vttbr(pgd_phys) | vmid; |
f7ed45be | 551 | |
f0cf47d9 | 552 | write_unlock(&kvm_vmid_lock); |
f7ed45be CD |
553 | } |
554 | ||
f7ed45be CD |
555 | static int kvm_vcpu_first_run_init(struct kvm_vcpu *vcpu) |
556 | { | |
05971120 | 557 | struct kvm *kvm = vcpu->kvm; |
41a54482 | 558 | int ret = 0; |
e1ba0207 | 559 | |
f7ed45be CD |
560 | if (likely(vcpu->arch.has_run_once)) |
561 | return 0; | |
562 | ||
563 | vcpu->arch.has_run_once = true; | |
aa024c2f | 564 | |
61bbe380 CD |
565 | if (likely(irqchip_in_kernel(kvm))) { |
566 | /* | |
567 | * Map the VGIC hardware resources before running a vcpu the | |
568 | * first time on this VM. | |
569 | */ | |
570 | if (unlikely(!vgic_ready(kvm))) { | |
571 | ret = kvm_vgic_map_resources(kvm); | |
572 | if (ret) | |
573 | return ret; | |
574 | } | |
575 | } else { | |
576 | /* | |
577 | * Tell the rest of the code that there are userspace irqchip | |
578 | * VMs in the wild. | |
579 | */ | |
580 | static_branch_inc(&userspace_irqchip_in_use); | |
01ac5e34 MZ |
581 | } |
582 | ||
d9e13977 | 583 | ret = kvm_timer_enable(vcpu); |
a2befacf CD |
584 | if (ret) |
585 | return ret; | |
586 | ||
587 | ret = kvm_arm_pmu_v3_enable(vcpu); | |
05971120 | 588 | |
41a54482 | 589 | return ret; |
f7ed45be CD |
590 | } |
591 | ||
c1426e4c EA |
592 | bool kvm_arch_intc_initialized(struct kvm *kvm) |
593 | { | |
594 | return vgic_initialized(kvm); | |
595 | } | |
596 | ||
b13216cf | 597 | void kvm_arm_halt_guest(struct kvm *kvm) |
3b92830a EA |
598 | { |
599 | int i; | |
600 | struct kvm_vcpu *vcpu; | |
601 | ||
602 | kvm_for_each_vcpu(i, vcpu, kvm) | |
603 | vcpu->arch.pause = true; | |
7b244e2b | 604 | kvm_make_all_cpus_request(kvm, KVM_REQ_SLEEP); |
3b92830a EA |
605 | } |
606 | ||
b13216cf | 607 | void kvm_arm_resume_guest(struct kvm *kvm) |
3b92830a EA |
608 | { |
609 | int i; | |
610 | struct kvm_vcpu *vcpu; | |
611 | ||
abd72296 CD |
612 | kvm_for_each_vcpu(i, vcpu, kvm) { |
613 | vcpu->arch.pause = false; | |
b3dae109 | 614 | swake_up_one(kvm_arch_vcpu_wq(vcpu)); |
abd72296 | 615 | } |
3b92830a EA |
616 | } |
617 | ||
7b244e2b | 618 | static void vcpu_req_sleep(struct kvm_vcpu *vcpu) |
aa024c2f | 619 | { |
8577370f | 620 | struct swait_queue_head *wq = kvm_arch_vcpu_wq(vcpu); |
aa024c2f | 621 | |
b3dae109 | 622 | swait_event_interruptible_exclusive(*wq, ((!vcpu->arch.power_off) && |
3b92830a | 623 | (!vcpu->arch.pause))); |
0592c005 | 624 | |
424c989b | 625 | if (vcpu->arch.power_off || vcpu->arch.pause) { |
0592c005 | 626 | /* Awaken to handle a signal, request we sleep again later. */ |
7b244e2b | 627 | kvm_make_request(KVM_REQ_SLEEP, vcpu); |
0592c005 | 628 | } |
aa024c2f MZ |
629 | } |
630 | ||
e8180dca AP |
631 | static int kvm_vcpu_initialized(struct kvm_vcpu *vcpu) |
632 | { | |
633 | return vcpu->arch.target >= 0; | |
634 | } | |
635 | ||
0592c005 AJ |
636 | static void check_vcpu_requests(struct kvm_vcpu *vcpu) |
637 | { | |
638 | if (kvm_request_pending(vcpu)) { | |
7b244e2b AJ |
639 | if (kvm_check_request(KVM_REQ_SLEEP, vcpu)) |
640 | vcpu_req_sleep(vcpu); | |
325f9c64 AJ |
641 | |
642 | /* | |
643 | * Clear IRQ_PENDING requests that were made to guarantee | |
644 | * that a VCPU sees new virtual interrupts. | |
645 | */ | |
646 | kvm_check_request(KVM_REQ_IRQ_PENDING, vcpu); | |
0592c005 AJ |
647 | } |
648 | } | |
649 | ||
f7ed45be CD |
650 | /** |
651 | * kvm_arch_vcpu_ioctl_run - the main VCPU run function to execute guest code | |
652 | * @vcpu: The VCPU pointer | |
653 | * @run: The kvm_run structure pointer used for userspace state exchange | |
654 | * | |
655 | * This function is called through the VCPU_RUN ioctl called from user space. It | |
656 | * will execute VM code in a loop until the time slice for the process is used | |
657 | * or some emulation is needed from user space in which case the function will | |
658 | * return with return value 0 and with the kvm_run structure filled in with the | |
659 | * required data for the requested emulation. | |
660 | */ | |
749cf76c CD |
661 | int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) |
662 | { | |
f7ed45be | 663 | int ret; |
f7ed45be | 664 | |
e8180dca | 665 | if (unlikely(!kvm_vcpu_initialized(vcpu))) |
f7ed45be CD |
666 | return -ENOEXEC; |
667 | ||
668 | ret = kvm_vcpu_first_run_init(vcpu); | |
669 | if (ret) | |
829a5863 | 670 | return ret; |
f7ed45be | 671 | |
45e96ea6 CD |
672 | if (run->exit_reason == KVM_EXIT_MMIO) { |
673 | ret = kvm_handle_mmio_return(vcpu, vcpu->run); | |
674 | if (ret) | |
829a5863 CD |
675 | return ret; |
676 | if (kvm_arm_handle_step_debug(vcpu, vcpu->run)) | |
677 | return 0; | |
45e96ea6 CD |
678 | } |
679 | ||
829a5863 CD |
680 | if (run->immediate_exit) |
681 | return -EINTR; | |
682 | ||
683 | vcpu_load(vcpu); | |
460df4c1 | 684 | |
20b7035c | 685 | kvm_sigset_activate(vcpu); |
f7ed45be CD |
686 | |
687 | ret = 1; | |
688 | run->exit_reason = KVM_EXIT_UNKNOWN; | |
689 | while (ret > 0) { | |
690 | /* | |
691 | * Check conditions before entering the guest | |
692 | */ | |
693 | cond_resched(); | |
694 | ||
695 | update_vttbr(vcpu->kvm); | |
696 | ||
0592c005 AJ |
697 | check_vcpu_requests(vcpu); |
698 | ||
abdf5843 MZ |
699 | /* |
700 | * Preparing the interrupts to be injected also | |
701 | * involves poking the GIC, which must be done in a | |
702 | * non-preemptible context. | |
703 | */ | |
1b3d546d | 704 | preempt_disable(); |
328e5664 | 705 | |
b02386eb | 706 | kvm_pmu_flush_hwstate(vcpu); |
328e5664 | 707 | |
f7ed45be CD |
708 | local_irq_disable(); |
709 | ||
abdf5843 MZ |
710 | kvm_vgic_flush_hwstate(vcpu); |
711 | ||
f7ed45be | 712 | /* |
61bbe380 CD |
713 | * Exit if we have a signal pending so that we can deliver the |
714 | * signal to user space. | |
f7ed45be | 715 | */ |
61bbe380 | 716 | if (signal_pending(current)) { |
f7ed45be CD |
717 | ret = -EINTR; |
718 | run->exit_reason = KVM_EXIT_INTR; | |
719 | } | |
720 | ||
61bbe380 CD |
721 | /* |
722 | * If we're using a userspace irqchip, then check if we need | |
723 | * to tell a userspace irqchip about timer or PMU level | |
724 | * changes and if so, exit to userspace (the actual level | |
725 | * state gets updated in kvm_timer_update_run and | |
726 | * kvm_pmu_update_run below). | |
727 | */ | |
728 | if (static_branch_unlikely(&userspace_irqchip_in_use)) { | |
729 | if (kvm_timer_should_notify_user(vcpu) || | |
730 | kvm_pmu_should_notify_user(vcpu)) { | |
731 | ret = -EINTR; | |
732 | run->exit_reason = KVM_EXIT_INTR; | |
733 | } | |
734 | } | |
735 | ||
6a6d73be AJ |
736 | /* |
737 | * Ensure we set mode to IN_GUEST_MODE after we disable | |
738 | * interrupts and before the final VCPU requests check. | |
739 | * See the comment in kvm_vcpu_exiting_guest_mode() and | |
740 | * Documentation/virtual/kvm/vcpu-requests.rst | |
741 | */ | |
742 | smp_store_mb(vcpu->mode, IN_GUEST_MODE); | |
743 | ||
101d3da0 | 744 | if (ret <= 0 || need_new_vmid_gen(vcpu->kvm) || |
424c989b | 745 | kvm_request_pending(vcpu)) { |
6a6d73be | 746 | vcpu->mode = OUTSIDE_GUEST_MODE; |
771621b0 | 747 | isb(); /* Ensure work in x_flush_hwstate is committed */ |
b02386eb | 748 | kvm_pmu_sync_hwstate(vcpu); |
61bbe380 CD |
749 | if (static_branch_unlikely(&userspace_irqchip_in_use)) |
750 | kvm_timer_sync_hwstate(vcpu); | |
1a89dd91 | 751 | kvm_vgic_sync_hwstate(vcpu); |
ee9bb9a1 | 752 | local_irq_enable(); |
abdf5843 | 753 | preempt_enable(); |
f7ed45be CD |
754 | continue; |
755 | } | |
756 | ||
56c7f5e7 AB |
757 | kvm_arm_setup_debug(vcpu); |
758 | ||
f7ed45be CD |
759 | /************************************************************** |
760 | * Enter the guest | |
761 | */ | |
762 | trace_kvm_entry(*vcpu_pc(vcpu)); | |
6edaa530 | 763 | guest_enter_irqoff(); |
f7ed45be | 764 | |
3f5c90b8 CD |
765 | if (has_vhe()) { |
766 | kvm_arm_vhe_guest_enter(); | |
767 | ret = kvm_vcpu_run_vhe(vcpu); | |
4f5abad9 | 768 | kvm_arm_vhe_guest_exit(); |
3f5c90b8 CD |
769 | } else { |
770 | ret = kvm_call_hyp(__kvm_vcpu_run_nvhe, vcpu); | |
771 | } | |
772 | ||
f7ed45be | 773 | vcpu->mode = OUTSIDE_GUEST_MODE; |
b19e6892 | 774 | vcpu->stat.exits++; |
1b3d546d CD |
775 | /* |
776 | * Back from guest | |
777 | *************************************************************/ | |
778 | ||
56c7f5e7 AB |
779 | kvm_arm_clear_debug(vcpu); |
780 | ||
ee9bb9a1 | 781 | /* |
b103cc3f | 782 | * We must sync the PMU state before the vgic state so |
ee9bb9a1 CD |
783 | * that the vgic can properly sample the updated state of the |
784 | * interrupt line. | |
785 | */ | |
786 | kvm_pmu_sync_hwstate(vcpu); | |
ee9bb9a1 | 787 | |
b103cc3f CD |
788 | /* |
789 | * Sync the vgic state before syncing the timer state because | |
790 | * the timer code needs to know if the virtual timer | |
791 | * interrupts are active. | |
792 | */ | |
ee9bb9a1 CD |
793 | kvm_vgic_sync_hwstate(vcpu); |
794 | ||
b103cc3f CD |
795 | /* |
796 | * Sync the timer hardware state before enabling interrupts as | |
797 | * we don't want vtimer interrupts to race with syncing the | |
798 | * timer virtual interrupt state. | |
799 | */ | |
61bbe380 CD |
800 | if (static_branch_unlikely(&userspace_irqchip_in_use)) |
801 | kvm_timer_sync_hwstate(vcpu); | |
b103cc3f | 802 | |
e6b673b7 DM |
803 | kvm_arch_vcpu_ctxsync_fp(vcpu); |
804 | ||
f7ed45be CD |
805 | /* |
806 | * We may have taken a host interrupt in HYP mode (ie | |
807 | * while executing the guest). This interrupt is still | |
808 | * pending, as we haven't serviced it yet! | |
809 | * | |
810 | * We're now back in SVC mode, with interrupts | |
811 | * disabled. Enabling the interrupts now will have | |
812 | * the effect of taking the interrupt again, in SVC | |
813 | * mode this time. | |
814 | */ | |
815 | local_irq_enable(); | |
816 | ||
817 | /* | |
6edaa530 | 818 | * We do local_irq_enable() before calling guest_exit() so |
1b3d546d CD |
819 | * that if a timer interrupt hits while running the guest we |
820 | * account that tick as being spent in the guest. We enable | |
6edaa530 | 821 | * preemption after calling guest_exit() so that if we get |
1b3d546d CD |
822 | * preempted we make sure ticks after that is not counted as |
823 | * guest time. | |
824 | */ | |
6edaa530 | 825 | guest_exit(); |
b5905dc1 | 826 | trace_kvm_exit(ret, kvm_vcpu_trap_get_class(vcpu), *vcpu_pc(vcpu)); |
1b3d546d | 827 | |
3368bd80 JM |
828 | /* Exit types that need handling before we can be preempted */ |
829 | handle_exit_early(vcpu, run, ret); | |
830 | ||
abdf5843 MZ |
831 | preempt_enable(); |
832 | ||
f7ed45be CD |
833 | ret = handle_exit(vcpu, run, ret); |
834 | } | |
835 | ||
d9e13977 | 836 | /* Tell userspace about in-kernel device output levels */ |
3dbbdf78 CD |
837 | if (unlikely(!irqchip_in_kernel(vcpu->kvm))) { |
838 | kvm_timer_update_run(vcpu); | |
839 | kvm_pmu_update_run(vcpu); | |
840 | } | |
d9e13977 | 841 | |
20b7035c JS |
842 | kvm_sigset_deactivate(vcpu); |
843 | ||
accb757d | 844 | vcpu_put(vcpu); |
f7ed45be | 845 | return ret; |
749cf76c CD |
846 | } |
847 | ||
86ce8535 CD |
848 | static int vcpu_interrupt_line(struct kvm_vcpu *vcpu, int number, bool level) |
849 | { | |
850 | int bit_index; | |
851 | bool set; | |
3df59d8d | 852 | unsigned long *hcr; |
86ce8535 CD |
853 | |
854 | if (number == KVM_ARM_IRQ_CPU_IRQ) | |
855 | bit_index = __ffs(HCR_VI); | |
856 | else /* KVM_ARM_IRQ_CPU_FIQ */ | |
857 | bit_index = __ffs(HCR_VF); | |
858 | ||
3df59d8d | 859 | hcr = vcpu_hcr(vcpu); |
86ce8535 | 860 | if (level) |
3df59d8d | 861 | set = test_and_set_bit(bit_index, hcr); |
86ce8535 | 862 | else |
3df59d8d | 863 | set = test_and_clear_bit(bit_index, hcr); |
86ce8535 CD |
864 | |
865 | /* | |
866 | * If we didn't change anything, no need to wake up or kick other CPUs | |
867 | */ | |
868 | if (set == level) | |
869 | return 0; | |
870 | ||
871 | /* | |
872 | * The vcpu irq_lines field was updated, wake up sleeping VCPUs and | |
873 | * trigger a world-switch round on the running physical CPU to set the | |
874 | * virtual IRQ/FIQ fields in the HCR appropriately. | |
875 | */ | |
325f9c64 | 876 | kvm_make_request(KVM_REQ_IRQ_PENDING, vcpu); |
86ce8535 CD |
877 | kvm_vcpu_kick(vcpu); |
878 | ||
879 | return 0; | |
880 | } | |
881 | ||
79558f11 AG |
882 | int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level, |
883 | bool line_status) | |
86ce8535 CD |
884 | { |
885 | u32 irq = irq_level->irq; | |
886 | unsigned int irq_type, vcpu_idx, irq_num; | |
887 | int nrcpus = atomic_read(&kvm->online_vcpus); | |
888 | struct kvm_vcpu *vcpu = NULL; | |
889 | bool level = irq_level->level; | |
890 | ||
891 | irq_type = (irq >> KVM_ARM_IRQ_TYPE_SHIFT) & KVM_ARM_IRQ_TYPE_MASK; | |
892 | vcpu_idx = (irq >> KVM_ARM_IRQ_VCPU_SHIFT) & KVM_ARM_IRQ_VCPU_MASK; | |
893 | irq_num = (irq >> KVM_ARM_IRQ_NUM_SHIFT) & KVM_ARM_IRQ_NUM_MASK; | |
894 | ||
895 | trace_kvm_irq_line(irq_type, vcpu_idx, irq_num, irq_level->level); | |
896 | ||
5863c2ce MZ |
897 | switch (irq_type) { |
898 | case KVM_ARM_IRQ_TYPE_CPU: | |
899 | if (irqchip_in_kernel(kvm)) | |
900 | return -ENXIO; | |
86ce8535 | 901 | |
5863c2ce MZ |
902 | if (vcpu_idx >= nrcpus) |
903 | return -EINVAL; | |
86ce8535 | 904 | |
5863c2ce MZ |
905 | vcpu = kvm_get_vcpu(kvm, vcpu_idx); |
906 | if (!vcpu) | |
907 | return -EINVAL; | |
86ce8535 | 908 | |
5863c2ce MZ |
909 | if (irq_num > KVM_ARM_IRQ_CPU_FIQ) |
910 | return -EINVAL; | |
911 | ||
912 | return vcpu_interrupt_line(vcpu, irq_num, level); | |
913 | case KVM_ARM_IRQ_TYPE_PPI: | |
914 | if (!irqchip_in_kernel(kvm)) | |
915 | return -ENXIO; | |
916 | ||
917 | if (vcpu_idx >= nrcpus) | |
918 | return -EINVAL; | |
919 | ||
920 | vcpu = kvm_get_vcpu(kvm, vcpu_idx); | |
921 | if (!vcpu) | |
922 | return -EINVAL; | |
923 | ||
924 | if (irq_num < VGIC_NR_SGIS || irq_num >= VGIC_NR_PRIVATE_IRQS) | |
925 | return -EINVAL; | |
86ce8535 | 926 | |
cb3f0ad8 | 927 | return kvm_vgic_inject_irq(kvm, vcpu->vcpu_id, irq_num, level, NULL); |
5863c2ce MZ |
928 | case KVM_ARM_IRQ_TYPE_SPI: |
929 | if (!irqchip_in_kernel(kvm)) | |
930 | return -ENXIO; | |
931 | ||
fd1d0ddf | 932 | if (irq_num < VGIC_NR_PRIVATE_IRQS) |
5863c2ce MZ |
933 | return -EINVAL; |
934 | ||
cb3f0ad8 | 935 | return kvm_vgic_inject_irq(kvm, 0, irq_num, level, NULL); |
5863c2ce MZ |
936 | } |
937 | ||
938 | return -EINVAL; | |
86ce8535 CD |
939 | } |
940 | ||
f7fa034d CD |
941 | static int kvm_vcpu_set_target(struct kvm_vcpu *vcpu, |
942 | const struct kvm_vcpu_init *init) | |
943 | { | |
944 | unsigned int i; | |
945 | int phys_target = kvm_target_cpu(); | |
946 | ||
947 | if (init->target != phys_target) | |
948 | return -EINVAL; | |
949 | ||
950 | /* | |
951 | * Secondary and subsequent calls to KVM_ARM_VCPU_INIT must | |
952 | * use the same target. | |
953 | */ | |
954 | if (vcpu->arch.target != -1 && vcpu->arch.target != init->target) | |
955 | return -EINVAL; | |
956 | ||
957 | /* -ENOENT for unknown features, -EINVAL for invalid combinations. */ | |
958 | for (i = 0; i < sizeof(init->features) * 8; i++) { | |
959 | bool set = (init->features[i / 32] & (1 << (i % 32))); | |
960 | ||
961 | if (set && i >= KVM_VCPU_MAX_FEATURES) | |
962 | return -ENOENT; | |
963 | ||
964 | /* | |
965 | * Secondary and subsequent calls to KVM_ARM_VCPU_INIT must | |
966 | * use the same feature set. | |
967 | */ | |
968 | if (vcpu->arch.target != -1 && i < KVM_VCPU_MAX_FEATURES && | |
969 | test_bit(i, vcpu->arch.features) != set) | |
970 | return -EINVAL; | |
971 | ||
972 | if (set) | |
973 | set_bit(i, vcpu->arch.features); | |
974 | } | |
975 | ||
976 | vcpu->arch.target = phys_target; | |
977 | ||
978 | /* Now we know what it is, we can reset it. */ | |
979 | return kvm_reset_vcpu(vcpu); | |
980 | } | |
981 | ||
982 | ||
478a8237 CD |
983 | static int kvm_arch_vcpu_ioctl_vcpu_init(struct kvm_vcpu *vcpu, |
984 | struct kvm_vcpu_init *init) | |
985 | { | |
986 | int ret; | |
987 | ||
988 | ret = kvm_vcpu_set_target(vcpu, init); | |
989 | if (ret) | |
990 | return ret; | |
991 | ||
957db105 CD |
992 | /* |
993 | * Ensure a rebooted VM will fault in RAM pages and detect if the | |
994 | * guest MMU is turned off and flush the caches as needed. | |
995 | */ | |
996 | if (vcpu->arch.has_run_once) | |
997 | stage2_unmap_vm(vcpu->kvm); | |
998 | ||
b856a591 CD |
999 | vcpu_reset_hcr(vcpu); |
1000 | ||
478a8237 | 1001 | /* |
3781528e | 1002 | * Handle the "start in power-off" case. |
478a8237 | 1003 | */ |
03f1d4c1 | 1004 | if (test_bit(KVM_ARM_VCPU_POWER_OFF, vcpu->arch.features)) |
424c989b | 1005 | vcpu_power_off(vcpu); |
3ad8b3de | 1006 | else |
3781528e | 1007 | vcpu->arch.power_off = false; |
478a8237 CD |
1008 | |
1009 | return 0; | |
1010 | } | |
1011 | ||
f577f6c2 SZ |
1012 | static int kvm_arm_vcpu_set_attr(struct kvm_vcpu *vcpu, |
1013 | struct kvm_device_attr *attr) | |
1014 | { | |
1015 | int ret = -ENXIO; | |
1016 | ||
1017 | switch (attr->group) { | |
1018 | default: | |
bb0c70bc | 1019 | ret = kvm_arm_vcpu_arch_set_attr(vcpu, attr); |
f577f6c2 SZ |
1020 | break; |
1021 | } | |
1022 | ||
1023 | return ret; | |
1024 | } | |
1025 | ||
1026 | static int kvm_arm_vcpu_get_attr(struct kvm_vcpu *vcpu, | |
1027 | struct kvm_device_attr *attr) | |
1028 | { | |
1029 | int ret = -ENXIO; | |
1030 | ||
1031 | switch (attr->group) { | |
1032 | default: | |
bb0c70bc | 1033 | ret = kvm_arm_vcpu_arch_get_attr(vcpu, attr); |
f577f6c2 SZ |
1034 | break; |
1035 | } | |
1036 | ||
1037 | return ret; | |
1038 | } | |
1039 | ||
1040 | static int kvm_arm_vcpu_has_attr(struct kvm_vcpu *vcpu, | |
1041 | struct kvm_device_attr *attr) | |
1042 | { | |
1043 | int ret = -ENXIO; | |
1044 | ||
1045 | switch (attr->group) { | |
1046 | default: | |
bb0c70bc | 1047 | ret = kvm_arm_vcpu_arch_has_attr(vcpu, attr); |
f577f6c2 SZ |
1048 | break; |
1049 | } | |
1050 | ||
1051 | return ret; | |
1052 | } | |
1053 | ||
539aee0e JM |
1054 | static int kvm_arm_vcpu_get_events(struct kvm_vcpu *vcpu, |
1055 | struct kvm_vcpu_events *events) | |
1056 | { | |
1057 | memset(events, 0, sizeof(*events)); | |
1058 | ||
1059 | return __kvm_arm_vcpu_get_events(vcpu, events); | |
1060 | } | |
1061 | ||
1062 | static int kvm_arm_vcpu_set_events(struct kvm_vcpu *vcpu, | |
1063 | struct kvm_vcpu_events *events) | |
1064 | { | |
1065 | int i; | |
1066 | ||
1067 | /* check whether the reserved field is zero */ | |
1068 | for (i = 0; i < ARRAY_SIZE(events->reserved); i++) | |
1069 | if (events->reserved[i]) | |
1070 | return -EINVAL; | |
1071 | ||
1072 | /* check whether the pad field is zero */ | |
1073 | for (i = 0; i < ARRAY_SIZE(events->exception.pad); i++) | |
1074 | if (events->exception.pad[i]) | |
1075 | return -EINVAL; | |
1076 | ||
1077 | return __kvm_arm_vcpu_set_events(vcpu, events); | |
1078 | } | |
539aee0e | 1079 | |
749cf76c CD |
1080 | long kvm_arch_vcpu_ioctl(struct file *filp, |
1081 | unsigned int ioctl, unsigned long arg) | |
1082 | { | |
1083 | struct kvm_vcpu *vcpu = filp->private_data; | |
1084 | void __user *argp = (void __user *)arg; | |
f577f6c2 | 1085 | struct kvm_device_attr attr; |
9b062471 CD |
1086 | long r; |
1087 | ||
749cf76c CD |
1088 | switch (ioctl) { |
1089 | case KVM_ARM_VCPU_INIT: { | |
1090 | struct kvm_vcpu_init init; | |
1091 | ||
9b062471 | 1092 | r = -EFAULT; |
749cf76c | 1093 | if (copy_from_user(&init, argp, sizeof(init))) |
9b062471 | 1094 | break; |
749cf76c | 1095 | |
9b062471 CD |
1096 | r = kvm_arch_vcpu_ioctl_vcpu_init(vcpu, &init); |
1097 | break; | |
749cf76c CD |
1098 | } |
1099 | case KVM_SET_ONE_REG: | |
1100 | case KVM_GET_ONE_REG: { | |
1101 | struct kvm_one_reg reg; | |
e8180dca | 1102 | |
9b062471 | 1103 | r = -ENOEXEC; |
e8180dca | 1104 | if (unlikely(!kvm_vcpu_initialized(vcpu))) |
9b062471 | 1105 | break; |
e8180dca | 1106 | |
9b062471 | 1107 | r = -EFAULT; |
749cf76c | 1108 | if (copy_from_user(®, argp, sizeof(reg))) |
9b062471 CD |
1109 | break; |
1110 | ||
749cf76c | 1111 | if (ioctl == KVM_SET_ONE_REG) |
9b062471 | 1112 | r = kvm_arm_set_reg(vcpu, ®); |
749cf76c | 1113 | else |
9b062471 CD |
1114 | r = kvm_arm_get_reg(vcpu, ®); |
1115 | break; | |
749cf76c CD |
1116 | } |
1117 | case KVM_GET_REG_LIST: { | |
1118 | struct kvm_reg_list __user *user_list = argp; | |
1119 | struct kvm_reg_list reg_list; | |
1120 | unsigned n; | |
1121 | ||
9b062471 | 1122 | r = -ENOEXEC; |
e8180dca | 1123 | if (unlikely(!kvm_vcpu_initialized(vcpu))) |
9b062471 | 1124 | break; |
e8180dca | 1125 | |
9b062471 | 1126 | r = -EFAULT; |
749cf76c | 1127 | if (copy_from_user(®_list, user_list, sizeof(reg_list))) |
9b062471 | 1128 | break; |
749cf76c CD |
1129 | n = reg_list.n; |
1130 | reg_list.n = kvm_arm_num_regs(vcpu); | |
1131 | if (copy_to_user(user_list, ®_list, sizeof(reg_list))) | |
9b062471 CD |
1132 | break; |
1133 | r = -E2BIG; | |
749cf76c | 1134 | if (n < reg_list.n) |
9b062471 CD |
1135 | break; |
1136 | r = kvm_arm_copy_reg_indices(vcpu, user_list->reg); | |
1137 | break; | |
749cf76c | 1138 | } |
f577f6c2 | 1139 | case KVM_SET_DEVICE_ATTR: { |
9b062471 | 1140 | r = -EFAULT; |
f577f6c2 | 1141 | if (copy_from_user(&attr, argp, sizeof(attr))) |
9b062471 CD |
1142 | break; |
1143 | r = kvm_arm_vcpu_set_attr(vcpu, &attr); | |
1144 | break; | |
f577f6c2 SZ |
1145 | } |
1146 | case KVM_GET_DEVICE_ATTR: { | |
9b062471 | 1147 | r = -EFAULT; |
f577f6c2 | 1148 | if (copy_from_user(&attr, argp, sizeof(attr))) |
9b062471 CD |
1149 | break; |
1150 | r = kvm_arm_vcpu_get_attr(vcpu, &attr); | |
1151 | break; | |
f577f6c2 SZ |
1152 | } |
1153 | case KVM_HAS_DEVICE_ATTR: { | |
9b062471 | 1154 | r = -EFAULT; |
f577f6c2 | 1155 | if (copy_from_user(&attr, argp, sizeof(attr))) |
9b062471 CD |
1156 | break; |
1157 | r = kvm_arm_vcpu_has_attr(vcpu, &attr); | |
1158 | break; | |
f577f6c2 | 1159 | } |
b7b27fac DG |
1160 | case KVM_GET_VCPU_EVENTS: { |
1161 | struct kvm_vcpu_events events; | |
1162 | ||
1163 | if (kvm_arm_vcpu_get_events(vcpu, &events)) | |
1164 | return -EINVAL; | |
1165 | ||
1166 | if (copy_to_user(argp, &events, sizeof(events))) | |
1167 | return -EFAULT; | |
1168 | ||
1169 | return 0; | |
1170 | } | |
1171 | case KVM_SET_VCPU_EVENTS: { | |
1172 | struct kvm_vcpu_events events; | |
1173 | ||
1174 | if (copy_from_user(&events, argp, sizeof(events))) | |
1175 | return -EFAULT; | |
1176 | ||
1177 | return kvm_arm_vcpu_set_events(vcpu, &events); | |
1178 | } | |
749cf76c | 1179 | default: |
9b062471 | 1180 | r = -EINVAL; |
749cf76c | 1181 | } |
9b062471 | 1182 | |
9b062471 | 1183 | return r; |
749cf76c CD |
1184 | } |
1185 | ||
53c810c3 MS |
1186 | /** |
1187 | * kvm_vm_ioctl_get_dirty_log - get and clear the log of dirty pages in a slot | |
1188 | * @kvm: kvm instance | |
1189 | * @log: slot id and address to which we copy the log | |
1190 | * | |
1191 | * Steps 1-4 below provide general overview of dirty page logging. See | |
1192 | * kvm_get_dirty_log_protect() function description for additional details. | |
1193 | * | |
1194 | * We call kvm_get_dirty_log_protect() to handle steps 1-3, upon return we | |
1195 | * always flush the TLB (step 4) even if previous step failed and the dirty | |
1196 | * bitmap may be corrupt. Regardless of previous outcome the KVM logging API | |
1197 | * does not preclude user space subsequent dirty log read. Flushing TLB ensures | |
1198 | * writes will be marked dirty for next log read. | |
1199 | * | |
1200 | * 1. Take a snapshot of the bit and clear it if needed. | |
1201 | * 2. Write protect the corresponding page. | |
1202 | * 3. Copy the snapshot to the userspace. | |
1203 | * 4. Flush TLB's if needed. | |
1204 | */ | |
749cf76c CD |
1205 | int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log) |
1206 | { | |
53c810c3 MS |
1207 | bool is_dirty = false; |
1208 | int r; | |
1209 | ||
1210 | mutex_lock(&kvm->slots_lock); | |
1211 | ||
1212 | r = kvm_get_dirty_log_protect(kvm, log, &is_dirty); | |
1213 | ||
1214 | if (is_dirty) | |
1215 | kvm_flush_remote_tlbs(kvm); | |
1216 | ||
1217 | mutex_unlock(&kvm->slots_lock); | |
1218 | return r; | |
749cf76c CD |
1219 | } |
1220 | ||
3401d546 CD |
1221 | static int kvm_vm_ioctl_set_device_addr(struct kvm *kvm, |
1222 | struct kvm_arm_device_addr *dev_addr) | |
1223 | { | |
330690cd CD |
1224 | unsigned long dev_id, type; |
1225 | ||
1226 | dev_id = (dev_addr->id & KVM_ARM_DEVICE_ID_MASK) >> | |
1227 | KVM_ARM_DEVICE_ID_SHIFT; | |
1228 | type = (dev_addr->id & KVM_ARM_DEVICE_TYPE_MASK) >> | |
1229 | KVM_ARM_DEVICE_TYPE_SHIFT; | |
1230 | ||
1231 | switch (dev_id) { | |
1232 | case KVM_ARM_DEVICE_VGIC_V2: | |
c7da6fa4 PF |
1233 | if (!vgic_present) |
1234 | return -ENXIO; | |
ce01e4e8 | 1235 | return kvm_vgic_addr(kvm, type, &dev_addr->addr, true); |
330690cd CD |
1236 | default: |
1237 | return -ENODEV; | |
1238 | } | |
3401d546 CD |
1239 | } |
1240 | ||
749cf76c CD |
1241 | long kvm_arch_vm_ioctl(struct file *filp, |
1242 | unsigned int ioctl, unsigned long arg) | |
1243 | { | |
3401d546 CD |
1244 | struct kvm *kvm = filp->private_data; |
1245 | void __user *argp = (void __user *)arg; | |
1246 | ||
1247 | switch (ioctl) { | |
5863c2ce | 1248 | case KVM_CREATE_IRQCHIP: { |
a28ebea2 | 1249 | int ret; |
c7da6fa4 PF |
1250 | if (!vgic_present) |
1251 | return -ENXIO; | |
a28ebea2 CD |
1252 | mutex_lock(&kvm->lock); |
1253 | ret = kvm_vgic_create(kvm, KVM_DEV_TYPE_ARM_VGIC_V2); | |
1254 | mutex_unlock(&kvm->lock); | |
1255 | return ret; | |
5863c2ce | 1256 | } |
3401d546 CD |
1257 | case KVM_ARM_SET_DEVICE_ADDR: { |
1258 | struct kvm_arm_device_addr dev_addr; | |
1259 | ||
1260 | if (copy_from_user(&dev_addr, argp, sizeof(dev_addr))) | |
1261 | return -EFAULT; | |
1262 | return kvm_vm_ioctl_set_device_addr(kvm, &dev_addr); | |
1263 | } | |
42c4e0c7 AP |
1264 | case KVM_ARM_PREFERRED_TARGET: { |
1265 | int err; | |
1266 | struct kvm_vcpu_init init; | |
1267 | ||
1268 | err = kvm_vcpu_preferred_target(&init); | |
1269 | if (err) | |
1270 | return err; | |
1271 | ||
1272 | if (copy_to_user(argp, &init, sizeof(init))) | |
1273 | return -EFAULT; | |
1274 | ||
1275 | return 0; | |
1276 | } | |
3401d546 CD |
1277 | default: |
1278 | return -EINVAL; | |
1279 | } | |
749cf76c CD |
1280 | } |
1281 | ||
d157f4a5 | 1282 | static void cpu_init_hyp_mode(void *dummy) |
342cd0ab | 1283 | { |
dac288f7 | 1284 | phys_addr_t pgd_ptr; |
342cd0ab CD |
1285 | unsigned long hyp_stack_ptr; |
1286 | unsigned long stack_page; | |
1287 | unsigned long vector_ptr; | |
1288 | ||
1289 | /* Switch from the HYP stub to our own HYP init vector */ | |
5a677ce0 | 1290 | __hyp_set_vectors(kvm_get_idmap_vector()); |
342cd0ab | 1291 | |
dac288f7 | 1292 | pgd_ptr = kvm_mmu_get_httbr(); |
1436c1aa | 1293 | stack_page = __this_cpu_read(kvm_arm_hyp_stack_page); |
342cd0ab | 1294 | hyp_stack_ptr = stack_page + PAGE_SIZE; |
6840bdd7 | 1295 | vector_ptr = (unsigned long)kvm_get_hyp_vector(); |
342cd0ab | 1296 | |
12fda812 | 1297 | __cpu_init_hyp_mode(pgd_ptr, hyp_stack_ptr, vector_ptr); |
35a2491a | 1298 | __cpu_init_stage2(); |
56c7f5e7 AB |
1299 | |
1300 | kvm_arm_init_debug(); | |
342cd0ab CD |
1301 | } |
1302 | ||
47eb3cba MZ |
1303 | static void cpu_hyp_reset(void) |
1304 | { | |
1305 | if (!is_kernel_in_hyp_mode()) | |
1306 | __hyp_reset_vectors(); | |
1307 | } | |
1308 | ||
5f5560b1 JM |
1309 | static void cpu_hyp_reinit(void) |
1310 | { | |
47eb3cba MZ |
1311 | cpu_hyp_reset(); |
1312 | ||
5f5560b1 JM |
1313 | if (is_kernel_in_hyp_mode()) { |
1314 | /* | |
67f69197 | 1315 | * __cpu_init_stage2() is safe to call even if the PM |
5f5560b1 JM |
1316 | * event was cancelled before the CPU was reset. |
1317 | */ | |
67f69197 | 1318 | __cpu_init_stage2(); |
02d50cda | 1319 | kvm_timer_init_vhe(); |
5f5560b1 | 1320 | } else { |
47eb3cba | 1321 | cpu_init_hyp_mode(NULL); |
5f5560b1 | 1322 | } |
5b0d2cc2 CD |
1323 | |
1324 | if (vgic_present) | |
1325 | kvm_vgic_init_cpu_hardware(); | |
5f5560b1 JM |
1326 | } |
1327 | ||
67f69197 AT |
1328 | static void _kvm_arch_hardware_enable(void *discard) |
1329 | { | |
1330 | if (!__this_cpu_read(kvm_arm_hardware_enabled)) { | |
5f5560b1 | 1331 | cpu_hyp_reinit(); |
67f69197 | 1332 | __this_cpu_write(kvm_arm_hardware_enabled, 1); |
d157f4a5 | 1333 | } |
67f69197 | 1334 | } |
d157f4a5 | 1335 | |
67f69197 AT |
1336 | int kvm_arch_hardware_enable(void) |
1337 | { | |
1338 | _kvm_arch_hardware_enable(NULL); | |
1339 | return 0; | |
342cd0ab CD |
1340 | } |
1341 | ||
67f69197 AT |
1342 | static void _kvm_arch_hardware_disable(void *discard) |
1343 | { | |
1344 | if (__this_cpu_read(kvm_arm_hardware_enabled)) { | |
1345 | cpu_hyp_reset(); | |
1346 | __this_cpu_write(kvm_arm_hardware_enabled, 0); | |
1347 | } | |
1348 | } | |
1349 | ||
1350 | void kvm_arch_hardware_disable(void) | |
1351 | { | |
1352 | _kvm_arch_hardware_disable(NULL); | |
1353 | } | |
d157f4a5 | 1354 | |
1fcf7ce0 LP |
1355 | #ifdef CONFIG_CPU_PM |
1356 | static int hyp_init_cpu_pm_notifier(struct notifier_block *self, | |
1357 | unsigned long cmd, | |
1358 | void *v) | |
1359 | { | |
67f69197 AT |
1360 | /* |
1361 | * kvm_arm_hardware_enabled is left with its old value over | |
1362 | * PM_ENTER->PM_EXIT. It is used to indicate PM_EXIT should | |
1363 | * re-enable hyp. | |
1364 | */ | |
1365 | switch (cmd) { | |
1366 | case CPU_PM_ENTER: | |
1367 | if (__this_cpu_read(kvm_arm_hardware_enabled)) | |
1368 | /* | |
1369 | * don't update kvm_arm_hardware_enabled here | |
1370 | * so that the hardware will be re-enabled | |
1371 | * when we resume. See below. | |
1372 | */ | |
1373 | cpu_hyp_reset(); | |
1374 | ||
1fcf7ce0 | 1375 | return NOTIFY_OK; |
58d6b15e | 1376 | case CPU_PM_ENTER_FAILED: |
67f69197 AT |
1377 | case CPU_PM_EXIT: |
1378 | if (__this_cpu_read(kvm_arm_hardware_enabled)) | |
1379 | /* The hardware was enabled before suspend. */ | |
1380 | cpu_hyp_reinit(); | |
1fcf7ce0 | 1381 | |
67f69197 AT |
1382 | return NOTIFY_OK; |
1383 | ||
1384 | default: | |
1385 | return NOTIFY_DONE; | |
1386 | } | |
1fcf7ce0 LP |
1387 | } |
1388 | ||
1389 | static struct notifier_block hyp_init_cpu_pm_nb = { | |
1390 | .notifier_call = hyp_init_cpu_pm_notifier, | |
1391 | }; | |
1392 | ||
1393 | static void __init hyp_cpu_pm_init(void) | |
1394 | { | |
1395 | cpu_pm_register_notifier(&hyp_init_cpu_pm_nb); | |
1396 | } | |
06a71a24 SH |
1397 | static void __init hyp_cpu_pm_exit(void) |
1398 | { | |
1399 | cpu_pm_unregister_notifier(&hyp_init_cpu_pm_nb); | |
1400 | } | |
1fcf7ce0 LP |
1401 | #else |
1402 | static inline void hyp_cpu_pm_init(void) | |
1403 | { | |
1404 | } | |
06a71a24 SH |
1405 | static inline void hyp_cpu_pm_exit(void) |
1406 | { | |
1407 | } | |
1fcf7ce0 LP |
1408 | #endif |
1409 | ||
1e947bad MZ |
1410 | static int init_common_resources(void) |
1411 | { | |
61349937 VM |
1412 | /* set size of VMID supported by CPU */ |
1413 | kvm_vmid_bits = kvm_get_vmid_bits(); | |
1414 | kvm_info("%d-bit VMID\n", kvm_vmid_bits); | |
1415 | ||
1e947bad MZ |
1416 | return 0; |
1417 | } | |
1418 | ||
1419 | static int init_subsystems(void) | |
1420 | { | |
67f69197 | 1421 | int err = 0; |
1e947bad | 1422 | |
5f5560b1 | 1423 | /* |
67f69197 | 1424 | * Enable hardware so that subsystem initialisation can access EL2. |
5f5560b1 | 1425 | */ |
67f69197 | 1426 | on_each_cpu(_kvm_arch_hardware_enable, NULL, 1); |
5f5560b1 JM |
1427 | |
1428 | /* | |
1429 | * Register CPU lower-power notifier | |
1430 | */ | |
1431 | hyp_cpu_pm_init(); | |
1432 | ||
1e947bad MZ |
1433 | /* |
1434 | * Init HYP view of VGIC | |
1435 | */ | |
1436 | err = kvm_vgic_hyp_init(); | |
1437 | switch (err) { | |
1438 | case 0: | |
1439 | vgic_present = true; | |
1440 | break; | |
1441 | case -ENODEV: | |
1442 | case -ENXIO: | |
1443 | vgic_present = false; | |
67f69197 | 1444 | err = 0; |
1e947bad MZ |
1445 | break; |
1446 | default: | |
67f69197 | 1447 | goto out; |
1e947bad MZ |
1448 | } |
1449 | ||
1450 | /* | |
1451 | * Init HYP architected timer support | |
1452 | */ | |
f384dcfe | 1453 | err = kvm_timer_hyp_init(vgic_present); |
1e947bad | 1454 | if (err) |
67f69197 | 1455 | goto out; |
1e947bad MZ |
1456 | |
1457 | kvm_perf_init(); | |
1458 | kvm_coproc_table_init(); | |
1459 | ||
67f69197 AT |
1460 | out: |
1461 | on_each_cpu(_kvm_arch_hardware_disable, NULL, 1); | |
1462 | ||
1463 | return err; | |
1e947bad MZ |
1464 | } |
1465 | ||
1466 | static void teardown_hyp_mode(void) | |
1467 | { | |
1468 | int cpu; | |
1469 | ||
1e947bad MZ |
1470 | free_hyp_pgds(); |
1471 | for_each_possible_cpu(cpu) | |
1472 | free_page(per_cpu(kvm_arm_hyp_stack_page, cpu)); | |
06a71a24 | 1473 | hyp_cpu_pm_exit(); |
1e947bad MZ |
1474 | } |
1475 | ||
342cd0ab CD |
1476 | /** |
1477 | * Inits Hyp-mode on all online CPUs | |
1478 | */ | |
1479 | static int init_hyp_mode(void) | |
1480 | { | |
342cd0ab CD |
1481 | int cpu; |
1482 | int err = 0; | |
1483 | ||
1484 | /* | |
1485 | * Allocate Hyp PGD and setup Hyp identity mapping | |
1486 | */ | |
1487 | err = kvm_mmu_init(); | |
1488 | if (err) | |
1489 | goto out_err; | |
1490 | ||
342cd0ab CD |
1491 | /* |
1492 | * Allocate stack pages for Hypervisor-mode | |
1493 | */ | |
1494 | for_each_possible_cpu(cpu) { | |
1495 | unsigned long stack_page; | |
1496 | ||
1497 | stack_page = __get_free_page(GFP_KERNEL); | |
1498 | if (!stack_page) { | |
1499 | err = -ENOMEM; | |
1e947bad | 1500 | goto out_err; |
342cd0ab CD |
1501 | } |
1502 | ||
1503 | per_cpu(kvm_arm_hyp_stack_page, cpu) = stack_page; | |
1504 | } | |
1505 | ||
342cd0ab CD |
1506 | /* |
1507 | * Map the Hyp-code called directly from the host | |
1508 | */ | |
588ab3f9 | 1509 | err = create_hyp_mappings(kvm_ksym_ref(__hyp_text_start), |
59002705 | 1510 | kvm_ksym_ref(__hyp_text_end), PAGE_HYP_EXEC); |
342cd0ab CD |
1511 | if (err) { |
1512 | kvm_err("Cannot map world-switch code\n"); | |
1e947bad | 1513 | goto out_err; |
342cd0ab CD |
1514 | } |
1515 | ||
a0bf9776 | 1516 | err = create_hyp_mappings(kvm_ksym_ref(__start_rodata), |
74a6b888 | 1517 | kvm_ksym_ref(__end_rodata), PAGE_HYP_RO); |
910917bb MZ |
1518 | if (err) { |
1519 | kvm_err("Cannot map rodata section\n"); | |
c8ea0395 MZ |
1520 | goto out_err; |
1521 | } | |
1522 | ||
1523 | err = create_hyp_mappings(kvm_ksym_ref(__bss_start), | |
1524 | kvm_ksym_ref(__bss_stop), PAGE_HYP_RO); | |
1525 | if (err) { | |
1526 | kvm_err("Cannot map bss section\n"); | |
1e947bad | 1527 | goto out_err; |
910917bb MZ |
1528 | } |
1529 | ||
6840bdd7 MZ |
1530 | err = kvm_map_vectors(); |
1531 | if (err) { | |
1532 | kvm_err("Cannot map vectors\n"); | |
1533 | goto out_err; | |
1534 | } | |
1535 | ||
342cd0ab CD |
1536 | /* |
1537 | * Map the Hyp stack pages | |
1538 | */ | |
1539 | for_each_possible_cpu(cpu) { | |
1540 | char *stack_page = (char *)per_cpu(kvm_arm_hyp_stack_page, cpu); | |
c8dddecd MZ |
1541 | err = create_hyp_mappings(stack_page, stack_page + PAGE_SIZE, |
1542 | PAGE_HYP); | |
342cd0ab CD |
1543 | |
1544 | if (err) { | |
1545 | kvm_err("Cannot map hyp stack\n"); | |
1e947bad | 1546 | goto out_err; |
342cd0ab CD |
1547 | } |
1548 | } | |
1549 | ||
342cd0ab | 1550 | for_each_possible_cpu(cpu) { |
3de50da6 | 1551 | kvm_cpu_context_t *cpu_ctxt; |
342cd0ab | 1552 | |
36989e7f | 1553 | cpu_ctxt = per_cpu_ptr(&kvm_host_cpu_state, cpu); |
c8dddecd | 1554 | err = create_hyp_mappings(cpu_ctxt, cpu_ctxt + 1, PAGE_HYP); |
342cd0ab CD |
1555 | |
1556 | if (err) { | |
3de50da6 | 1557 | kvm_err("Cannot map host CPU state: %d\n", err); |
1e947bad | 1558 | goto out_err; |
342cd0ab CD |
1559 | } |
1560 | } | |
1561 | ||
55e3748e MZ |
1562 | err = hyp_map_aux_data(); |
1563 | if (err) | |
1564 | kvm_err("Cannot map host auxilary data: %d\n", err); | |
1565 | ||
342cd0ab | 1566 | return 0; |
1e947bad | 1567 | |
342cd0ab | 1568 | out_err: |
1e947bad | 1569 | teardown_hyp_mode(); |
342cd0ab CD |
1570 | kvm_err("error initializing Hyp mode: %d\n", err); |
1571 | return err; | |
1572 | } | |
1573 | ||
d4e071ce AP |
1574 | static void check_kvm_target_cpu(void *ret) |
1575 | { | |
1576 | *(int *)ret = kvm_target_cpu(); | |
1577 | } | |
1578 | ||
4429fc64 AP |
1579 | struct kvm_vcpu *kvm_mpidr_to_vcpu(struct kvm *kvm, unsigned long mpidr) |
1580 | { | |
1581 | struct kvm_vcpu *vcpu; | |
1582 | int i; | |
1583 | ||
1584 | mpidr &= MPIDR_HWID_BITMASK; | |
1585 | kvm_for_each_vcpu(i, vcpu, kvm) { | |
1586 | if (mpidr == kvm_vcpu_get_mpidr_aff(vcpu)) | |
1587 | return vcpu; | |
1588 | } | |
1589 | return NULL; | |
1590 | } | |
1591 | ||
2412405b EA |
1592 | bool kvm_arch_has_irq_bypass(void) |
1593 | { | |
1594 | return true; | |
1595 | } | |
1596 | ||
1597 | int kvm_arch_irq_bypass_add_producer(struct irq_bypass_consumer *cons, | |
1598 | struct irq_bypass_producer *prod) | |
1599 | { | |
1600 | struct kvm_kernel_irqfd *irqfd = | |
1601 | container_of(cons, struct kvm_kernel_irqfd, consumer); | |
1602 | ||
196b1364 MZ |
1603 | return kvm_vgic_v4_set_forwarding(irqfd->kvm, prod->irq, |
1604 | &irqfd->irq_entry); | |
2412405b EA |
1605 | } |
1606 | void kvm_arch_irq_bypass_del_producer(struct irq_bypass_consumer *cons, | |
1607 | struct irq_bypass_producer *prod) | |
1608 | { | |
1609 | struct kvm_kernel_irqfd *irqfd = | |
1610 | container_of(cons, struct kvm_kernel_irqfd, consumer); | |
1611 | ||
196b1364 MZ |
1612 | kvm_vgic_v4_unset_forwarding(irqfd->kvm, prod->irq, |
1613 | &irqfd->irq_entry); | |
2412405b EA |
1614 | } |
1615 | ||
1616 | void kvm_arch_irq_bypass_stop(struct irq_bypass_consumer *cons) | |
1617 | { | |
1618 | struct kvm_kernel_irqfd *irqfd = | |
1619 | container_of(cons, struct kvm_kernel_irqfd, consumer); | |
1620 | ||
1621 | kvm_arm_halt_guest(irqfd->kvm); | |
1622 | } | |
1623 | ||
1624 | void kvm_arch_irq_bypass_start(struct irq_bypass_consumer *cons) | |
1625 | { | |
1626 | struct kvm_kernel_irqfd *irqfd = | |
1627 | container_of(cons, struct kvm_kernel_irqfd, consumer); | |
1628 | ||
1629 | kvm_arm_resume_guest(irqfd->kvm); | |
1630 | } | |
1631 | ||
342cd0ab CD |
1632 | /** |
1633 | * Initialize Hyp-mode and memory mappings on all CPUs. | |
1634 | */ | |
749cf76c CD |
1635 | int kvm_arch_init(void *opaque) |
1636 | { | |
342cd0ab | 1637 | int err; |
d4e071ce | 1638 | int ret, cpu; |
fe7d7b03 | 1639 | bool in_hyp_mode; |
342cd0ab CD |
1640 | |
1641 | if (!is_hyp_mode_available()) { | |
58d0d19a | 1642 | kvm_info("HYP mode not available\n"); |
342cd0ab CD |
1643 | return -ENODEV; |
1644 | } | |
1645 | ||
85acda3b DM |
1646 | if (!kvm_arch_check_sve_has_vhe()) { |
1647 | kvm_pr_unimpl("SVE system without VHE unsupported. Broken cpu?"); | |
1648 | return -ENODEV; | |
1649 | } | |
1650 | ||
d4e071ce AP |
1651 | for_each_online_cpu(cpu) { |
1652 | smp_call_function_single(cpu, check_kvm_target_cpu, &ret, 1); | |
1653 | if (ret < 0) { | |
1654 | kvm_err("Error, CPU %d not supported!\n", cpu); | |
1655 | return -ENODEV; | |
1656 | } | |
342cd0ab CD |
1657 | } |
1658 | ||
1e947bad | 1659 | err = init_common_resources(); |
342cd0ab | 1660 | if (err) |
1e947bad | 1661 | return err; |
342cd0ab | 1662 | |
fe7d7b03 JT |
1663 | in_hyp_mode = is_kernel_in_hyp_mode(); |
1664 | ||
1665 | if (!in_hyp_mode) { | |
1e947bad | 1666 | err = init_hyp_mode(); |
fe7d7b03 JT |
1667 | if (err) |
1668 | goto out_err; | |
1669 | } | |
8146875d | 1670 | |
1e947bad MZ |
1671 | err = init_subsystems(); |
1672 | if (err) | |
1673 | goto out_hyp; | |
1fcf7ce0 | 1674 | |
fe7d7b03 JT |
1675 | if (in_hyp_mode) |
1676 | kvm_info("VHE mode initialized successfully\n"); | |
1677 | else | |
1678 | kvm_info("Hyp mode initialized successfully\n"); | |
1679 | ||
749cf76c | 1680 | return 0; |
1e947bad MZ |
1681 | |
1682 | out_hyp: | |
fe7d7b03 JT |
1683 | if (!in_hyp_mode) |
1684 | teardown_hyp_mode(); | |
342cd0ab CD |
1685 | out_err: |
1686 | return err; | |
749cf76c CD |
1687 | } |
1688 | ||
1689 | /* NOP: Compiling as a module not supported */ | |
1690 | void kvm_arch_exit(void) | |
1691 | { | |
210552c1 | 1692 | kvm_perf_teardown(); |
749cf76c CD |
1693 | } |
1694 | ||
1695 | static int arm_init(void) | |
1696 | { | |
1697 | int rc = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE); | |
1698 | return rc; | |
1699 | } | |
1700 | ||
1701 | module_init(arm_init); |