]>
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 | ||
d157f4a5 | 19 | #include <linux/cpu.h> |
749cf76c CD |
20 | #include <linux/errno.h> |
21 | #include <linux/err.h> | |
22 | #include <linux/kvm_host.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/vmalloc.h> | |
25 | #include <linux/fs.h> | |
26 | #include <linux/mman.h> | |
27 | #include <linux/sched.h> | |
86ce8535 | 28 | #include <linux/kvm.h> |
749cf76c CD |
29 | #include <trace/events/kvm.h> |
30 | ||
31 | #define CREATE_TRACE_POINTS | |
32 | #include "trace.h" | |
33 | ||
749cf76c CD |
34 | #include <asm/uaccess.h> |
35 | #include <asm/ptrace.h> | |
36 | #include <asm/mman.h> | |
342cd0ab | 37 | #include <asm/tlbflush.h> |
5b3e5e5b | 38 | #include <asm/cacheflush.h> |
342cd0ab CD |
39 | #include <asm/virt.h> |
40 | #include <asm/kvm_arm.h> | |
41 | #include <asm/kvm_asm.h> | |
42 | #include <asm/kvm_mmu.h> | |
f7ed45be | 43 | #include <asm/kvm_emulate.h> |
5b3e5e5b | 44 | #include <asm/kvm_coproc.h> |
aa024c2f | 45 | #include <asm/kvm_psci.h> |
749cf76c CD |
46 | |
47 | #ifdef REQUIRES_VIRT | |
48 | __asm__(".arch_extension virt"); | |
49 | #endif | |
50 | ||
342cd0ab | 51 | static DEFINE_PER_CPU(unsigned long, kvm_arm_hyp_stack_page); |
3de50da6 | 52 | static kvm_cpu_context_t __percpu *kvm_host_cpu_state; |
342cd0ab CD |
53 | static unsigned long hyp_default_vectors; |
54 | ||
1638a12d MZ |
55 | /* Per-CPU variable containing the currently running vcpu. */ |
56 | static DEFINE_PER_CPU(struct kvm_vcpu *, kvm_arm_running_vcpu); | |
57 | ||
f7ed45be CD |
58 | /* The VMID used in the VTTBR */ |
59 | static atomic64_t kvm_vmid_gen = ATOMIC64_INIT(1); | |
60 | static u8 kvm_next_vmid; | |
61 | static DEFINE_SPINLOCK(kvm_vmid_lock); | |
342cd0ab | 62 | |
1a89dd91 MZ |
63 | static bool vgic_present; |
64 | ||
1638a12d MZ |
65 | static void kvm_arm_set_running_vcpu(struct kvm_vcpu *vcpu) |
66 | { | |
67 | BUG_ON(preemptible()); | |
68 | __get_cpu_var(kvm_arm_running_vcpu) = vcpu; | |
69 | } | |
70 | ||
71 | /** | |
72 | * kvm_arm_get_running_vcpu - get the vcpu running on the current CPU. | |
73 | * Must be called from non-preemptible context | |
74 | */ | |
75 | struct kvm_vcpu *kvm_arm_get_running_vcpu(void) | |
76 | { | |
77 | BUG_ON(preemptible()); | |
78 | return __get_cpu_var(kvm_arm_running_vcpu); | |
79 | } | |
80 | ||
81 | /** | |
82 | * kvm_arm_get_running_vcpus - get the per-CPU array of currently running vcpus. | |
83 | */ | |
84 | struct kvm_vcpu __percpu **kvm_get_running_vcpus(void) | |
85 | { | |
86 | return &kvm_arm_running_vcpu; | |
87 | } | |
88 | ||
749cf76c CD |
89 | int kvm_arch_hardware_enable(void *garbage) |
90 | { | |
91 | return 0; | |
92 | } | |
93 | ||
94 | int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu) | |
95 | { | |
96 | return kvm_vcpu_exiting_guest_mode(vcpu) == IN_GUEST_MODE; | |
97 | } | |
98 | ||
99 | void kvm_arch_hardware_disable(void *garbage) | |
100 | { | |
101 | } | |
102 | ||
103 | int kvm_arch_hardware_setup(void) | |
104 | { | |
105 | return 0; | |
106 | } | |
107 | ||
108 | void kvm_arch_hardware_unsetup(void) | |
109 | { | |
110 | } | |
111 | ||
112 | void kvm_arch_check_processor_compat(void *rtn) | |
113 | { | |
114 | *(int *)rtn = 0; | |
115 | } | |
116 | ||
117 | void kvm_arch_sync_events(struct kvm *kvm) | |
118 | { | |
119 | } | |
120 | ||
d5d8184d CD |
121 | /** |
122 | * kvm_arch_init_vm - initializes a VM data structure | |
123 | * @kvm: pointer to the KVM struct | |
124 | */ | |
749cf76c CD |
125 | int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) |
126 | { | |
d5d8184d CD |
127 | int ret = 0; |
128 | ||
749cf76c CD |
129 | if (type) |
130 | return -EINVAL; | |
131 | ||
d5d8184d CD |
132 | ret = kvm_alloc_stage2_pgd(kvm); |
133 | if (ret) | |
134 | goto out_fail_alloc; | |
135 | ||
136 | ret = create_hyp_mappings(kvm, kvm + 1); | |
137 | if (ret) | |
138 | goto out_free_stage2_pgd; | |
139 | ||
140 | /* Mark the initial VMID generation invalid */ | |
141 | kvm->arch.vmid_gen = 0; | |
142 | ||
143 | return ret; | |
144 | out_free_stage2_pgd: | |
145 | kvm_free_stage2_pgd(kvm); | |
146 | out_fail_alloc: | |
147 | return ret; | |
749cf76c CD |
148 | } |
149 | ||
150 | int kvm_arch_vcpu_fault(struct kvm_vcpu *vcpu, struct vm_fault *vmf) | |
151 | { | |
152 | return VM_FAULT_SIGBUS; | |
153 | } | |
154 | ||
155 | void kvm_arch_free_memslot(struct kvm_memory_slot *free, | |
156 | struct kvm_memory_slot *dont) | |
157 | { | |
158 | } | |
159 | ||
160 | int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages) | |
161 | { | |
162 | return 0; | |
163 | } | |
164 | ||
d5d8184d CD |
165 | /** |
166 | * kvm_arch_destroy_vm - destroy the VM data structure | |
167 | * @kvm: pointer to the KVM struct | |
168 | */ | |
749cf76c CD |
169 | void kvm_arch_destroy_vm(struct kvm *kvm) |
170 | { | |
171 | int i; | |
172 | ||
d5d8184d CD |
173 | kvm_free_stage2_pgd(kvm); |
174 | ||
749cf76c CD |
175 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
176 | if (kvm->vcpus[i]) { | |
177 | kvm_arch_vcpu_free(kvm->vcpus[i]); | |
178 | kvm->vcpus[i] = NULL; | |
179 | } | |
180 | } | |
181 | } | |
182 | ||
183 | int kvm_dev_ioctl_check_extension(long ext) | |
184 | { | |
185 | int r; | |
186 | switch (ext) { | |
1a89dd91 MZ |
187 | case KVM_CAP_IRQCHIP: |
188 | r = vgic_present; | |
189 | break; | |
749cf76c CD |
190 | case KVM_CAP_USER_MEMORY: |
191 | case KVM_CAP_SYNC_MMU: | |
192 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: | |
193 | case KVM_CAP_ONE_REG: | |
aa024c2f | 194 | case KVM_CAP_ARM_PSCI: |
749cf76c CD |
195 | r = 1; |
196 | break; | |
197 | case KVM_CAP_COALESCED_MMIO: | |
198 | r = KVM_COALESCED_MMIO_PAGE_OFFSET; | |
199 | break; | |
3401d546 CD |
200 | case KVM_CAP_ARM_SET_DEVICE_ADDR: |
201 | r = 1; | |
749cf76c CD |
202 | case KVM_CAP_NR_VCPUS: |
203 | r = num_online_cpus(); | |
204 | break; | |
205 | case KVM_CAP_MAX_VCPUS: | |
206 | r = KVM_MAX_VCPUS; | |
207 | break; | |
208 | default: | |
17b1e31f | 209 | r = kvm_arch_dev_ioctl_check_extension(ext); |
749cf76c CD |
210 | break; |
211 | } | |
212 | return r; | |
213 | } | |
214 | ||
215 | long kvm_arch_dev_ioctl(struct file *filp, | |
216 | unsigned int ioctl, unsigned long arg) | |
217 | { | |
218 | return -EINVAL; | |
219 | } | |
220 | ||
749cf76c CD |
221 | int kvm_arch_prepare_memory_region(struct kvm *kvm, |
222 | struct kvm_memory_slot *memslot, | |
7b6195a9 TY |
223 | struct kvm_userspace_memory_region *mem, |
224 | enum kvm_mr_change change) | |
749cf76c CD |
225 | { |
226 | return 0; | |
227 | } | |
228 | ||
229 | void kvm_arch_commit_memory_region(struct kvm *kvm, | |
230 | struct kvm_userspace_memory_region *mem, | |
8482644a TY |
231 | const struct kvm_memory_slot *old, |
232 | enum kvm_mr_change change) | |
749cf76c CD |
233 | { |
234 | } | |
235 | ||
236 | void kvm_arch_flush_shadow_all(struct kvm *kvm) | |
237 | { | |
238 | } | |
239 | ||
240 | void kvm_arch_flush_shadow_memslot(struct kvm *kvm, | |
241 | struct kvm_memory_slot *slot) | |
242 | { | |
243 | } | |
244 | ||
245 | struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm, unsigned int id) | |
246 | { | |
247 | int err; | |
248 | struct kvm_vcpu *vcpu; | |
249 | ||
250 | vcpu = kmem_cache_zalloc(kvm_vcpu_cache, GFP_KERNEL); | |
251 | if (!vcpu) { | |
252 | err = -ENOMEM; | |
253 | goto out; | |
254 | } | |
255 | ||
256 | err = kvm_vcpu_init(vcpu, kvm, id); | |
257 | if (err) | |
258 | goto free_vcpu; | |
259 | ||
d5d8184d CD |
260 | err = create_hyp_mappings(vcpu, vcpu + 1); |
261 | if (err) | |
262 | goto vcpu_uninit; | |
263 | ||
749cf76c | 264 | return vcpu; |
d5d8184d CD |
265 | vcpu_uninit: |
266 | kvm_vcpu_uninit(vcpu); | |
749cf76c CD |
267 | free_vcpu: |
268 | kmem_cache_free(kvm_vcpu_cache, vcpu); | |
269 | out: | |
270 | return ERR_PTR(err); | |
271 | } | |
272 | ||
273 | int kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu) | |
274 | { | |
275 | return 0; | |
276 | } | |
277 | ||
278 | void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu) | |
279 | { | |
d5d8184d | 280 | kvm_mmu_free_memory_caches(vcpu); |
967f8427 | 281 | kvm_timer_vcpu_terminate(vcpu); |
d5d8184d | 282 | kmem_cache_free(kvm_vcpu_cache, vcpu); |
749cf76c CD |
283 | } |
284 | ||
285 | void kvm_arch_vcpu_destroy(struct kvm_vcpu *vcpu) | |
286 | { | |
287 | kvm_arch_vcpu_free(vcpu); | |
288 | } | |
289 | ||
290 | int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu) | |
291 | { | |
292 | return 0; | |
293 | } | |
294 | ||
749cf76c CD |
295 | int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu) |
296 | { | |
1a89dd91 MZ |
297 | int ret; |
298 | ||
f7ed45be CD |
299 | /* Force users to call KVM_ARM_VCPU_INIT */ |
300 | vcpu->arch.target = -1; | |
1a89dd91 MZ |
301 | |
302 | /* Set up VGIC */ | |
303 | ret = kvm_vgic_vcpu_init(vcpu); | |
304 | if (ret) | |
305 | return ret; | |
306 | ||
967f8427 MZ |
307 | /* Set up the timer */ |
308 | kvm_timer_vcpu_init(vcpu); | |
309 | ||
749cf76c CD |
310 | return 0; |
311 | } | |
312 | ||
313 | void kvm_arch_vcpu_uninit(struct kvm_vcpu *vcpu) | |
314 | { | |
315 | } | |
316 | ||
317 | void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu) | |
318 | { | |
86ce8535 | 319 | vcpu->cpu = cpu; |
3de50da6 | 320 | vcpu->arch.host_cpu_context = this_cpu_ptr(kvm_host_cpu_state); |
5b3e5e5b CD |
321 | |
322 | /* | |
323 | * Check whether this vcpu requires the cache to be flushed on | |
324 | * this physical CPU. This is a consequence of doing dcache | |
325 | * operations by set/way on this vcpu. We do it here to be in | |
326 | * a non-preemptible section. | |
327 | */ | |
328 | if (cpumask_test_and_clear_cpu(cpu, &vcpu->arch.require_dcache_flush)) | |
329 | flush_cache_all(); /* We'd really want v7_flush_dcache_all() */ | |
1638a12d MZ |
330 | |
331 | kvm_arm_set_running_vcpu(vcpu); | |
749cf76c CD |
332 | } |
333 | ||
334 | void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu) | |
335 | { | |
1638a12d | 336 | kvm_arm_set_running_vcpu(NULL); |
749cf76c CD |
337 | } |
338 | ||
339 | int kvm_arch_vcpu_ioctl_set_guest_debug(struct kvm_vcpu *vcpu, | |
340 | struct kvm_guest_debug *dbg) | |
341 | { | |
342 | return -EINVAL; | |
343 | } | |
344 | ||
345 | ||
346 | int kvm_arch_vcpu_ioctl_get_mpstate(struct kvm_vcpu *vcpu, | |
347 | struct kvm_mp_state *mp_state) | |
348 | { | |
349 | return -EINVAL; | |
350 | } | |
351 | ||
352 | int kvm_arch_vcpu_ioctl_set_mpstate(struct kvm_vcpu *vcpu, | |
353 | struct kvm_mp_state *mp_state) | |
354 | { | |
355 | return -EINVAL; | |
356 | } | |
357 | ||
5b3e5e5b CD |
358 | /** |
359 | * kvm_arch_vcpu_runnable - determine if the vcpu can be scheduled | |
360 | * @v: The VCPU pointer | |
361 | * | |
362 | * If the guest CPU is not waiting for interrupts or an interrupt line is | |
363 | * asserted, the CPU is by definition runnable. | |
364 | */ | |
749cf76c CD |
365 | int kvm_arch_vcpu_runnable(struct kvm_vcpu *v) |
366 | { | |
1a89dd91 | 367 | return !!v->arch.irq_lines || kvm_vgic_vcpu_pending_irq(v); |
749cf76c CD |
368 | } |
369 | ||
f7ed45be CD |
370 | /* Just ensure a guest exit from a particular CPU */ |
371 | static void exit_vm_noop(void *info) | |
372 | { | |
373 | } | |
374 | ||
375 | void force_vm_exit(const cpumask_t *mask) | |
376 | { | |
377 | smp_call_function_many(mask, exit_vm_noop, NULL, true); | |
378 | } | |
379 | ||
380 | /** | |
381 | * need_new_vmid_gen - check that the VMID is still valid | |
382 | * @kvm: The VM's VMID to checkt | |
383 | * | |
384 | * return true if there is a new generation of VMIDs being used | |
385 | * | |
386 | * The hardware supports only 256 values with the value zero reserved for the | |
387 | * host, so we check if an assigned value belongs to a previous generation, | |
388 | * which which requires us to assign a new value. If we're the first to use a | |
389 | * VMID for the new generation, we must flush necessary caches and TLBs on all | |
390 | * CPUs. | |
391 | */ | |
392 | static bool need_new_vmid_gen(struct kvm *kvm) | |
393 | { | |
394 | return unlikely(kvm->arch.vmid_gen != atomic64_read(&kvm_vmid_gen)); | |
395 | } | |
396 | ||
397 | /** | |
398 | * update_vttbr - Update the VTTBR with a valid VMID before the guest runs | |
399 | * @kvm The guest that we are about to run | |
400 | * | |
401 | * Called from kvm_arch_vcpu_ioctl_run before entering the guest to ensure the | |
402 | * VM has a valid VMID, otherwise assigns a new one and flushes corresponding | |
403 | * caches and TLBs. | |
404 | */ | |
405 | static void update_vttbr(struct kvm *kvm) | |
406 | { | |
407 | phys_addr_t pgd_phys; | |
408 | u64 vmid; | |
409 | ||
410 | if (!need_new_vmid_gen(kvm)) | |
411 | return; | |
412 | ||
413 | spin_lock(&kvm_vmid_lock); | |
414 | ||
415 | /* | |
416 | * We need to re-check the vmid_gen here to ensure that if another vcpu | |
417 | * already allocated a valid vmid for this vm, then this vcpu should | |
418 | * use the same vmid. | |
419 | */ | |
420 | if (!need_new_vmid_gen(kvm)) { | |
421 | spin_unlock(&kvm_vmid_lock); | |
422 | return; | |
423 | } | |
424 | ||
425 | /* First user of a new VMID generation? */ | |
426 | if (unlikely(kvm_next_vmid == 0)) { | |
427 | atomic64_inc(&kvm_vmid_gen); | |
428 | kvm_next_vmid = 1; | |
429 | ||
430 | /* | |
431 | * On SMP we know no other CPUs can use this CPU's or each | |
432 | * other's VMID after force_vm_exit returns since the | |
433 | * kvm_vmid_lock blocks them from reentry to the guest. | |
434 | */ | |
435 | force_vm_exit(cpu_all_mask); | |
436 | /* | |
437 | * Now broadcast TLB + ICACHE invalidation over the inner | |
438 | * shareable domain to make sure all data structures are | |
439 | * clean. | |
440 | */ | |
441 | kvm_call_hyp(__kvm_flush_vm_context); | |
442 | } | |
443 | ||
444 | kvm->arch.vmid_gen = atomic64_read(&kvm_vmid_gen); | |
445 | kvm->arch.vmid = kvm_next_vmid; | |
446 | kvm_next_vmid++; | |
447 | ||
448 | /* update vttbr to be used with the new vmid */ | |
449 | pgd_phys = virt_to_phys(kvm->arch.pgd); | |
450 | vmid = ((u64)(kvm->arch.vmid) << VTTBR_VMID_SHIFT) & VTTBR_VMID_MASK; | |
451 | kvm->arch.vttbr = pgd_phys & VTTBR_BADDR_MASK; | |
452 | kvm->arch.vttbr |= vmid; | |
453 | ||
454 | spin_unlock(&kvm_vmid_lock); | |
455 | } | |
456 | ||
f7ed45be CD |
457 | static int kvm_vcpu_first_run_init(struct kvm_vcpu *vcpu) |
458 | { | |
459 | if (likely(vcpu->arch.has_run_once)) | |
460 | return 0; | |
461 | ||
462 | vcpu->arch.has_run_once = true; | |
aa024c2f | 463 | |
01ac5e34 MZ |
464 | /* |
465 | * Initialize the VGIC before running a vcpu the first time on | |
466 | * this VM. | |
467 | */ | |
468 | if (irqchip_in_kernel(vcpu->kvm) && | |
469 | unlikely(!vgic_initialized(vcpu->kvm))) { | |
470 | int ret = kvm_vgic_init(vcpu->kvm); | |
471 | if (ret) | |
472 | return ret; | |
473 | } | |
474 | ||
aa024c2f MZ |
475 | /* |
476 | * Handle the "start in power-off" case by calling into the | |
477 | * PSCI code. | |
478 | */ | |
479 | if (test_and_clear_bit(KVM_ARM_VCPU_POWER_OFF, vcpu->arch.features)) { | |
480 | *vcpu_reg(vcpu, 0) = KVM_PSCI_FN_CPU_OFF; | |
481 | kvm_psci_call(vcpu); | |
482 | } | |
483 | ||
f7ed45be CD |
484 | return 0; |
485 | } | |
486 | ||
aa024c2f MZ |
487 | static void vcpu_pause(struct kvm_vcpu *vcpu) |
488 | { | |
489 | wait_queue_head_t *wq = kvm_arch_vcpu_wq(vcpu); | |
490 | ||
491 | wait_event_interruptible(*wq, !vcpu->arch.pause); | |
492 | } | |
493 | ||
f7ed45be CD |
494 | /** |
495 | * kvm_arch_vcpu_ioctl_run - the main VCPU run function to execute guest code | |
496 | * @vcpu: The VCPU pointer | |
497 | * @run: The kvm_run structure pointer used for userspace state exchange | |
498 | * | |
499 | * This function is called through the VCPU_RUN ioctl called from user space. It | |
500 | * will execute VM code in a loop until the time slice for the process is used | |
501 | * or some emulation is needed from user space in which case the function will | |
502 | * return with return value 0 and with the kvm_run structure filled in with the | |
503 | * required data for the requested emulation. | |
504 | */ | |
749cf76c CD |
505 | int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) |
506 | { | |
f7ed45be CD |
507 | int ret; |
508 | sigset_t sigsaved; | |
509 | ||
510 | /* Make sure they initialize the vcpu with KVM_ARM_VCPU_INIT */ | |
511 | if (unlikely(vcpu->arch.target < 0)) | |
512 | return -ENOEXEC; | |
513 | ||
514 | ret = kvm_vcpu_first_run_init(vcpu); | |
515 | if (ret) | |
516 | return ret; | |
517 | ||
45e96ea6 CD |
518 | if (run->exit_reason == KVM_EXIT_MMIO) { |
519 | ret = kvm_handle_mmio_return(vcpu, vcpu->run); | |
520 | if (ret) | |
521 | return ret; | |
522 | } | |
523 | ||
f7ed45be CD |
524 | if (vcpu->sigset_active) |
525 | sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); | |
526 | ||
527 | ret = 1; | |
528 | run->exit_reason = KVM_EXIT_UNKNOWN; | |
529 | while (ret > 0) { | |
530 | /* | |
531 | * Check conditions before entering the guest | |
532 | */ | |
533 | cond_resched(); | |
534 | ||
535 | update_vttbr(vcpu->kvm); | |
536 | ||
aa024c2f MZ |
537 | if (vcpu->arch.pause) |
538 | vcpu_pause(vcpu); | |
539 | ||
1a89dd91 | 540 | kvm_vgic_flush_hwstate(vcpu); |
c7e3ba64 | 541 | kvm_timer_flush_hwstate(vcpu); |
1a89dd91 | 542 | |
f7ed45be CD |
543 | local_irq_disable(); |
544 | ||
545 | /* | |
546 | * Re-check atomic conditions | |
547 | */ | |
548 | if (signal_pending(current)) { | |
549 | ret = -EINTR; | |
550 | run->exit_reason = KVM_EXIT_INTR; | |
551 | } | |
552 | ||
553 | if (ret <= 0 || need_new_vmid_gen(vcpu->kvm)) { | |
554 | local_irq_enable(); | |
c7e3ba64 | 555 | kvm_timer_sync_hwstate(vcpu); |
1a89dd91 | 556 | kvm_vgic_sync_hwstate(vcpu); |
f7ed45be CD |
557 | continue; |
558 | } | |
559 | ||
560 | /************************************************************** | |
561 | * Enter the guest | |
562 | */ | |
563 | trace_kvm_entry(*vcpu_pc(vcpu)); | |
564 | kvm_guest_enter(); | |
565 | vcpu->mode = IN_GUEST_MODE; | |
566 | ||
567 | ret = kvm_call_hyp(__kvm_vcpu_run, vcpu); | |
568 | ||
569 | vcpu->mode = OUTSIDE_GUEST_MODE; | |
5b3e5e5b | 570 | vcpu->arch.last_pcpu = smp_processor_id(); |
f7ed45be CD |
571 | kvm_guest_exit(); |
572 | trace_kvm_exit(*vcpu_pc(vcpu)); | |
573 | /* | |
574 | * We may have taken a host interrupt in HYP mode (ie | |
575 | * while executing the guest). This interrupt is still | |
576 | * pending, as we haven't serviced it yet! | |
577 | * | |
578 | * We're now back in SVC mode, with interrupts | |
579 | * disabled. Enabling the interrupts now will have | |
580 | * the effect of taking the interrupt again, in SVC | |
581 | * mode this time. | |
582 | */ | |
583 | local_irq_enable(); | |
584 | ||
585 | /* | |
586 | * Back from guest | |
587 | *************************************************************/ | |
588 | ||
c7e3ba64 | 589 | kvm_timer_sync_hwstate(vcpu); |
1a89dd91 MZ |
590 | kvm_vgic_sync_hwstate(vcpu); |
591 | ||
f7ed45be CD |
592 | ret = handle_exit(vcpu, run, ret); |
593 | } | |
594 | ||
595 | if (vcpu->sigset_active) | |
596 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
597 | return ret; | |
749cf76c CD |
598 | } |
599 | ||
86ce8535 CD |
600 | static int vcpu_interrupt_line(struct kvm_vcpu *vcpu, int number, bool level) |
601 | { | |
602 | int bit_index; | |
603 | bool set; | |
604 | unsigned long *ptr; | |
605 | ||
606 | if (number == KVM_ARM_IRQ_CPU_IRQ) | |
607 | bit_index = __ffs(HCR_VI); | |
608 | else /* KVM_ARM_IRQ_CPU_FIQ */ | |
609 | bit_index = __ffs(HCR_VF); | |
610 | ||
611 | ptr = (unsigned long *)&vcpu->arch.irq_lines; | |
612 | if (level) | |
613 | set = test_and_set_bit(bit_index, ptr); | |
614 | else | |
615 | set = test_and_clear_bit(bit_index, ptr); | |
616 | ||
617 | /* | |
618 | * If we didn't change anything, no need to wake up or kick other CPUs | |
619 | */ | |
620 | if (set == level) | |
621 | return 0; | |
622 | ||
623 | /* | |
624 | * The vcpu irq_lines field was updated, wake up sleeping VCPUs and | |
625 | * trigger a world-switch round on the running physical CPU to set the | |
626 | * virtual IRQ/FIQ fields in the HCR appropriately. | |
627 | */ | |
628 | kvm_vcpu_kick(vcpu); | |
629 | ||
630 | return 0; | |
631 | } | |
632 | ||
79558f11 AG |
633 | int kvm_vm_ioctl_irq_line(struct kvm *kvm, struct kvm_irq_level *irq_level, |
634 | bool line_status) | |
86ce8535 CD |
635 | { |
636 | u32 irq = irq_level->irq; | |
637 | unsigned int irq_type, vcpu_idx, irq_num; | |
638 | int nrcpus = atomic_read(&kvm->online_vcpus); | |
639 | struct kvm_vcpu *vcpu = NULL; | |
640 | bool level = irq_level->level; | |
641 | ||
642 | irq_type = (irq >> KVM_ARM_IRQ_TYPE_SHIFT) & KVM_ARM_IRQ_TYPE_MASK; | |
643 | vcpu_idx = (irq >> KVM_ARM_IRQ_VCPU_SHIFT) & KVM_ARM_IRQ_VCPU_MASK; | |
644 | irq_num = (irq >> KVM_ARM_IRQ_NUM_SHIFT) & KVM_ARM_IRQ_NUM_MASK; | |
645 | ||
646 | trace_kvm_irq_line(irq_type, vcpu_idx, irq_num, irq_level->level); | |
647 | ||
5863c2ce MZ |
648 | switch (irq_type) { |
649 | case KVM_ARM_IRQ_TYPE_CPU: | |
650 | if (irqchip_in_kernel(kvm)) | |
651 | return -ENXIO; | |
86ce8535 | 652 | |
5863c2ce MZ |
653 | if (vcpu_idx >= nrcpus) |
654 | return -EINVAL; | |
86ce8535 | 655 | |
5863c2ce MZ |
656 | vcpu = kvm_get_vcpu(kvm, vcpu_idx); |
657 | if (!vcpu) | |
658 | return -EINVAL; | |
86ce8535 | 659 | |
5863c2ce MZ |
660 | if (irq_num > KVM_ARM_IRQ_CPU_FIQ) |
661 | return -EINVAL; | |
662 | ||
663 | return vcpu_interrupt_line(vcpu, irq_num, level); | |
664 | case KVM_ARM_IRQ_TYPE_PPI: | |
665 | if (!irqchip_in_kernel(kvm)) | |
666 | return -ENXIO; | |
667 | ||
668 | if (vcpu_idx >= nrcpus) | |
669 | return -EINVAL; | |
670 | ||
671 | vcpu = kvm_get_vcpu(kvm, vcpu_idx); | |
672 | if (!vcpu) | |
673 | return -EINVAL; | |
674 | ||
675 | if (irq_num < VGIC_NR_SGIS || irq_num >= VGIC_NR_PRIVATE_IRQS) | |
676 | return -EINVAL; | |
86ce8535 | 677 | |
5863c2ce MZ |
678 | return kvm_vgic_inject_irq(kvm, vcpu->vcpu_id, irq_num, level); |
679 | case KVM_ARM_IRQ_TYPE_SPI: | |
680 | if (!irqchip_in_kernel(kvm)) | |
681 | return -ENXIO; | |
682 | ||
683 | if (irq_num < VGIC_NR_PRIVATE_IRQS || | |
684 | irq_num > KVM_ARM_IRQ_GIC_MAX) | |
685 | return -EINVAL; | |
686 | ||
687 | return kvm_vgic_inject_irq(kvm, 0, irq_num, level); | |
688 | } | |
689 | ||
690 | return -EINVAL; | |
86ce8535 CD |
691 | } |
692 | ||
749cf76c CD |
693 | long kvm_arch_vcpu_ioctl(struct file *filp, |
694 | unsigned int ioctl, unsigned long arg) | |
695 | { | |
696 | struct kvm_vcpu *vcpu = filp->private_data; | |
697 | void __user *argp = (void __user *)arg; | |
698 | ||
699 | switch (ioctl) { | |
700 | case KVM_ARM_VCPU_INIT: { | |
701 | struct kvm_vcpu_init init; | |
702 | ||
703 | if (copy_from_user(&init, argp, sizeof(init))) | |
704 | return -EFAULT; | |
705 | ||
706 | return kvm_vcpu_set_target(vcpu, &init); | |
707 | ||
708 | } | |
709 | case KVM_SET_ONE_REG: | |
710 | case KVM_GET_ONE_REG: { | |
711 | struct kvm_one_reg reg; | |
712 | if (copy_from_user(®, argp, sizeof(reg))) | |
713 | return -EFAULT; | |
714 | if (ioctl == KVM_SET_ONE_REG) | |
715 | return kvm_arm_set_reg(vcpu, ®); | |
716 | else | |
717 | return kvm_arm_get_reg(vcpu, ®); | |
718 | } | |
719 | case KVM_GET_REG_LIST: { | |
720 | struct kvm_reg_list __user *user_list = argp; | |
721 | struct kvm_reg_list reg_list; | |
722 | unsigned n; | |
723 | ||
724 | if (copy_from_user(®_list, user_list, sizeof(reg_list))) | |
725 | return -EFAULT; | |
726 | n = reg_list.n; | |
727 | reg_list.n = kvm_arm_num_regs(vcpu); | |
728 | if (copy_to_user(user_list, ®_list, sizeof(reg_list))) | |
729 | return -EFAULT; | |
730 | if (n < reg_list.n) | |
731 | return -E2BIG; | |
732 | return kvm_arm_copy_reg_indices(vcpu, user_list->reg); | |
733 | } | |
734 | default: | |
735 | return -EINVAL; | |
736 | } | |
737 | } | |
738 | ||
739 | int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, struct kvm_dirty_log *log) | |
740 | { | |
741 | return -EINVAL; | |
742 | } | |
743 | ||
3401d546 CD |
744 | static int kvm_vm_ioctl_set_device_addr(struct kvm *kvm, |
745 | struct kvm_arm_device_addr *dev_addr) | |
746 | { | |
330690cd CD |
747 | unsigned long dev_id, type; |
748 | ||
749 | dev_id = (dev_addr->id & KVM_ARM_DEVICE_ID_MASK) >> | |
750 | KVM_ARM_DEVICE_ID_SHIFT; | |
751 | type = (dev_addr->id & KVM_ARM_DEVICE_TYPE_MASK) >> | |
752 | KVM_ARM_DEVICE_TYPE_SHIFT; | |
753 | ||
754 | switch (dev_id) { | |
755 | case KVM_ARM_DEVICE_VGIC_V2: | |
756 | if (!vgic_present) | |
757 | return -ENXIO; | |
758 | return kvm_vgic_set_addr(kvm, type, dev_addr->addr); | |
759 | default: | |
760 | return -ENODEV; | |
761 | } | |
3401d546 CD |
762 | } |
763 | ||
749cf76c CD |
764 | long kvm_arch_vm_ioctl(struct file *filp, |
765 | unsigned int ioctl, unsigned long arg) | |
766 | { | |
3401d546 CD |
767 | struct kvm *kvm = filp->private_data; |
768 | void __user *argp = (void __user *)arg; | |
769 | ||
770 | switch (ioctl) { | |
5863c2ce MZ |
771 | case KVM_CREATE_IRQCHIP: { |
772 | if (vgic_present) | |
773 | return kvm_vgic_create(kvm); | |
774 | else | |
775 | return -ENXIO; | |
776 | } | |
3401d546 CD |
777 | case KVM_ARM_SET_DEVICE_ADDR: { |
778 | struct kvm_arm_device_addr dev_addr; | |
779 | ||
780 | if (copy_from_user(&dev_addr, argp, sizeof(dev_addr))) | |
781 | return -EFAULT; | |
782 | return kvm_vm_ioctl_set_device_addr(kvm, &dev_addr); | |
783 | } | |
784 | default: | |
785 | return -EINVAL; | |
786 | } | |
749cf76c CD |
787 | } |
788 | ||
d157f4a5 | 789 | static void cpu_init_hyp_mode(void *dummy) |
342cd0ab | 790 | { |
5a677ce0 | 791 | unsigned long long boot_pgd_ptr; |
342cd0ab | 792 | unsigned long long pgd_ptr; |
342cd0ab CD |
793 | unsigned long hyp_stack_ptr; |
794 | unsigned long stack_page; | |
795 | unsigned long vector_ptr; | |
796 | ||
797 | /* Switch from the HYP stub to our own HYP init vector */ | |
5a677ce0 | 798 | __hyp_set_vectors(kvm_get_idmap_vector()); |
342cd0ab | 799 | |
5a677ce0 | 800 | boot_pgd_ptr = (unsigned long long)kvm_mmu_get_boot_httbr(); |
342cd0ab | 801 | pgd_ptr = (unsigned long long)kvm_mmu_get_httbr(); |
342cd0ab CD |
802 | stack_page = __get_cpu_var(kvm_arm_hyp_stack_page); |
803 | hyp_stack_ptr = stack_page + PAGE_SIZE; | |
804 | vector_ptr = (unsigned long)__kvm_hyp_vector; | |
805 | ||
5a677ce0 | 806 | __cpu_init_hyp_mode(boot_pgd_ptr, pgd_ptr, hyp_stack_ptr, vector_ptr); |
342cd0ab CD |
807 | } |
808 | ||
d157f4a5 MZ |
809 | static int hyp_init_cpu_notify(struct notifier_block *self, |
810 | unsigned long action, void *cpu) | |
811 | { | |
812 | switch (action) { | |
813 | case CPU_STARTING: | |
814 | case CPU_STARTING_FROZEN: | |
815 | cpu_init_hyp_mode(NULL); | |
816 | break; | |
817 | } | |
818 | ||
819 | return NOTIFY_OK; | |
820 | } | |
821 | ||
822 | static struct notifier_block hyp_init_cpu_nb = { | |
823 | .notifier_call = hyp_init_cpu_notify, | |
824 | }; | |
825 | ||
342cd0ab CD |
826 | /** |
827 | * Inits Hyp-mode on all online CPUs | |
828 | */ | |
829 | static int init_hyp_mode(void) | |
830 | { | |
342cd0ab CD |
831 | int cpu; |
832 | int err = 0; | |
833 | ||
834 | /* | |
835 | * Allocate Hyp PGD and setup Hyp identity mapping | |
836 | */ | |
837 | err = kvm_mmu_init(); | |
838 | if (err) | |
839 | goto out_err; | |
840 | ||
841 | /* | |
842 | * It is probably enough to obtain the default on one | |
843 | * CPU. It's unlikely to be different on the others. | |
844 | */ | |
845 | hyp_default_vectors = __hyp_get_vectors(); | |
846 | ||
847 | /* | |
848 | * Allocate stack pages for Hypervisor-mode | |
849 | */ | |
850 | for_each_possible_cpu(cpu) { | |
851 | unsigned long stack_page; | |
852 | ||
853 | stack_page = __get_free_page(GFP_KERNEL); | |
854 | if (!stack_page) { | |
855 | err = -ENOMEM; | |
856 | goto out_free_stack_pages; | |
857 | } | |
858 | ||
859 | per_cpu(kvm_arm_hyp_stack_page, cpu) = stack_page; | |
860 | } | |
861 | ||
342cd0ab CD |
862 | /* |
863 | * Map the Hyp-code called directly from the host | |
864 | */ | |
865 | err = create_hyp_mappings(__kvm_hyp_code_start, __kvm_hyp_code_end); | |
866 | if (err) { | |
867 | kvm_err("Cannot map world-switch code\n"); | |
868 | goto out_free_mappings; | |
869 | } | |
870 | ||
871 | /* | |
872 | * Map the Hyp stack pages | |
873 | */ | |
874 | for_each_possible_cpu(cpu) { | |
875 | char *stack_page = (char *)per_cpu(kvm_arm_hyp_stack_page, cpu); | |
876 | err = create_hyp_mappings(stack_page, stack_page + PAGE_SIZE); | |
877 | ||
878 | if (err) { | |
879 | kvm_err("Cannot map hyp stack\n"); | |
880 | goto out_free_mappings; | |
881 | } | |
882 | } | |
883 | ||
884 | /* | |
3de50da6 | 885 | * Map the host CPU structures |
342cd0ab | 886 | */ |
3de50da6 MZ |
887 | kvm_host_cpu_state = alloc_percpu(kvm_cpu_context_t); |
888 | if (!kvm_host_cpu_state) { | |
342cd0ab | 889 | err = -ENOMEM; |
3de50da6 | 890 | kvm_err("Cannot allocate host CPU state\n"); |
342cd0ab CD |
891 | goto out_free_mappings; |
892 | } | |
893 | ||
894 | for_each_possible_cpu(cpu) { | |
3de50da6 | 895 | kvm_cpu_context_t *cpu_ctxt; |
342cd0ab | 896 | |
3de50da6 MZ |
897 | cpu_ctxt = per_cpu_ptr(kvm_host_cpu_state, cpu); |
898 | err = create_hyp_mappings(cpu_ctxt, cpu_ctxt + 1); | |
342cd0ab CD |
899 | |
900 | if (err) { | |
3de50da6 MZ |
901 | kvm_err("Cannot map host CPU state: %d\n", err); |
902 | goto out_free_context; | |
342cd0ab CD |
903 | } |
904 | } | |
905 | ||
d157f4a5 MZ |
906 | /* |
907 | * Execute the init code on each CPU. | |
908 | */ | |
909 | on_each_cpu(cpu_init_hyp_mode, NULL, 1); | |
910 | ||
1a89dd91 MZ |
911 | /* |
912 | * Init HYP view of VGIC | |
913 | */ | |
914 | err = kvm_vgic_hyp_init(); | |
915 | if (err) | |
3de50da6 | 916 | goto out_free_context; |
1a89dd91 | 917 | |
01ac5e34 MZ |
918 | #ifdef CONFIG_KVM_ARM_VGIC |
919 | vgic_present = true; | |
920 | #endif | |
921 | ||
967f8427 MZ |
922 | /* |
923 | * Init HYP architected timer support | |
924 | */ | |
925 | err = kvm_timer_hyp_init(); | |
926 | if (err) | |
927 | goto out_free_mappings; | |
928 | ||
d157f4a5 MZ |
929 | #ifndef CONFIG_HOTPLUG_CPU |
930 | free_boot_hyp_pgd(); | |
931 | #endif | |
932 | ||
210552c1 MZ |
933 | kvm_perf_init(); |
934 | ||
342cd0ab | 935 | kvm_info("Hyp mode initialized successfully\n"); |
210552c1 | 936 | |
342cd0ab | 937 | return 0; |
3de50da6 MZ |
938 | out_free_context: |
939 | free_percpu(kvm_host_cpu_state); | |
342cd0ab | 940 | out_free_mappings: |
4f728276 | 941 | free_hyp_pgds(); |
342cd0ab CD |
942 | out_free_stack_pages: |
943 | for_each_possible_cpu(cpu) | |
944 | free_page(per_cpu(kvm_arm_hyp_stack_page, cpu)); | |
945 | out_err: | |
946 | kvm_err("error initializing Hyp mode: %d\n", err); | |
947 | return err; | |
948 | } | |
949 | ||
950 | /** | |
951 | * Initialize Hyp-mode and memory mappings on all CPUs. | |
952 | */ | |
749cf76c CD |
953 | int kvm_arch_init(void *opaque) |
954 | { | |
342cd0ab CD |
955 | int err; |
956 | ||
957 | if (!is_hyp_mode_available()) { | |
958 | kvm_err("HYP mode not available\n"); | |
959 | return -ENODEV; | |
960 | } | |
961 | ||
962 | if (kvm_target_cpu() < 0) { | |
963 | kvm_err("Target CPU not supported!\n"); | |
964 | return -ENODEV; | |
965 | } | |
966 | ||
967 | err = init_hyp_mode(); | |
968 | if (err) | |
969 | goto out_err; | |
970 | ||
d157f4a5 MZ |
971 | err = register_cpu_notifier(&hyp_init_cpu_nb); |
972 | if (err) { | |
973 | kvm_err("Cannot register HYP init CPU notifier (%d)\n", err); | |
974 | goto out_err; | |
975 | } | |
976 | ||
5b3e5e5b | 977 | kvm_coproc_table_init(); |
749cf76c | 978 | return 0; |
342cd0ab CD |
979 | out_err: |
980 | return err; | |
749cf76c CD |
981 | } |
982 | ||
983 | /* NOP: Compiling as a module not supported */ | |
984 | void kvm_arch_exit(void) | |
985 | { | |
210552c1 | 986 | kvm_perf_teardown(); |
749cf76c CD |
987 | } |
988 | ||
989 | static int arm_init(void) | |
990 | { | |
991 | int rc = kvm_init(NULL, sizeof(struct kvm_vcpu), 0, THIS_MODULE); | |
992 | return rc; | |
993 | } | |
994 | ||
995 | module_init(arm_init); |