]>
Commit | Line | Data |
---|---|---|
6aa8b732 AK |
1 | /* |
2 | * Kernel-based Virtual Machine driver for Linux | |
3 | * | |
4 | * This module enables machines with Intel VT-x extensions to run virtual | |
5 | * machines without emulation or binary translation. | |
6 | * | |
7 | * Copyright (C) 2006 Qumranet, Inc. | |
9611c187 | 8 | * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
6aa8b732 AK |
9 | * |
10 | * Authors: | |
11 | * Avi Kivity <avi@qumranet.com> | |
12 | * Yaniv Kamay <yaniv@qumranet.com> | |
13 | * | |
14 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
15 | * the COPYING file in the top-level directory. | |
16 | * | |
17 | */ | |
18 | ||
e2174021 | 19 | #include "iodev.h" |
6aa8b732 | 20 | |
edf88417 | 21 | #include <linux/kvm_host.h> |
6aa8b732 AK |
22 | #include <linux/kvm.h> |
23 | #include <linux/module.h> | |
24 | #include <linux/errno.h> | |
6aa8b732 | 25 | #include <linux/percpu.h> |
6aa8b732 AK |
26 | #include <linux/mm.h> |
27 | #include <linux/miscdevice.h> | |
28 | #include <linux/vmalloc.h> | |
6aa8b732 | 29 | #include <linux/reboot.h> |
6aa8b732 AK |
30 | #include <linux/debugfs.h> |
31 | #include <linux/highmem.h> | |
32 | #include <linux/file.h> | |
59ae6c6b | 33 | #include <linux/sysdev.h> |
774c47f1 | 34 | #include <linux/cpu.h> |
e8edc6e0 | 35 | #include <linux/sched.h> |
d9e368d6 AK |
36 | #include <linux/cpumask.h> |
37 | #include <linux/smp.h> | |
d6d28168 | 38 | #include <linux/anon_inodes.h> |
04d2cc77 | 39 | #include <linux/profile.h> |
7aa81cc0 | 40 | #include <linux/kvm_para.h> |
6fc138d2 | 41 | #include <linux/pagemap.h> |
8d4e1288 | 42 | #include <linux/mman.h> |
35149e21 | 43 | #include <linux/swap.h> |
e56d532f | 44 | #include <linux/bitops.h> |
547de29e | 45 | #include <linux/spinlock.h> |
6ff5894c | 46 | #include <linux/compat.h> |
bc6678a3 | 47 | #include <linux/srcu.h> |
8f0b1ab6 | 48 | #include <linux/hugetlb.h> |
5a0e3ad6 | 49 | #include <linux/slab.h> |
6aa8b732 | 50 | |
e495606d | 51 | #include <asm/processor.h> |
e495606d AK |
52 | #include <asm/io.h> |
53 | #include <asm/uaccess.h> | |
3e021bf5 | 54 | #include <asm/pgtable.h> |
c8240bd6 | 55 | #include <asm-generic/bitops/le.h> |
6aa8b732 | 56 | |
5f94c174 | 57 | #include "coalesced_mmio.h" |
af585b92 | 58 | #include "async_pf.h" |
5f94c174 | 59 | |
229456fc MT |
60 | #define CREATE_TRACE_POINTS |
61 | #include <trace/events/kvm.h> | |
62 | ||
6aa8b732 AK |
63 | MODULE_AUTHOR("Qumranet"); |
64 | MODULE_LICENSE("GPL"); | |
65 | ||
fa40a821 MT |
66 | /* |
67 | * Ordering of locks: | |
68 | * | |
fae3a353 | 69 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
70 | */ |
71 | ||
e9b11c17 ZX |
72 | DEFINE_SPINLOCK(kvm_lock); |
73 | LIST_HEAD(vm_list); | |
133de902 | 74 | |
7f59f492 | 75 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
76 | static int kvm_usage_count = 0; |
77 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 78 | |
c16f862d RR |
79 | struct kmem_cache *kvm_vcpu_cache; |
80 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 81 | |
15ad7146 AK |
82 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
83 | ||
76f7c879 | 84 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 85 | |
bccf2150 AK |
86 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
87 | unsigned long arg); | |
10474ae8 AG |
88 | static int hardware_enable_all(void); |
89 | static void hardware_disable_all(void); | |
bccf2150 | 90 | |
e93f8a0f MT |
91 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
92 | ||
e8ba5d31 | 93 | static bool kvm_rebooting; |
4ecac3fd | 94 | |
54dee993 MT |
95 | static bool largepages_enabled = true; |
96 | ||
fa7bff8f GN |
97 | static struct page *hwpoison_page; |
98 | static pfn_t hwpoison_pfn; | |
bf998156 | 99 | |
edba23e5 GN |
100 | static struct page *fault_page; |
101 | static pfn_t fault_pfn; | |
102 | ||
c77fb9dc | 103 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 104 | { |
fc5659c8 JR |
105 | if (pfn_valid(pfn)) { |
106 | struct page *page = compound_head(pfn_to_page(pfn)); | |
107 | return PageReserved(page); | |
108 | } | |
cbff90a7 BAY |
109 | |
110 | return true; | |
111 | } | |
112 | ||
bccf2150 AK |
113 | /* |
114 | * Switches to specified vcpu, until a matching vcpu_put() | |
115 | */ | |
313a3dc7 | 116 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 117 | { |
15ad7146 AK |
118 | int cpu; |
119 | ||
bccf2150 | 120 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
121 | cpu = get_cpu(); |
122 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 123 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 124 | put_cpu(); |
6aa8b732 AK |
125 | } |
126 | ||
313a3dc7 | 127 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 128 | { |
15ad7146 | 129 | preempt_disable(); |
313a3dc7 | 130 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
131 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
132 | preempt_enable(); | |
6aa8b732 AK |
133 | mutex_unlock(&vcpu->mutex); |
134 | } | |
135 | ||
d9e368d6 AK |
136 | static void ack_flush(void *_completed) |
137 | { | |
d9e368d6 AK |
138 | } |
139 | ||
49846896 | 140 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 141 | { |
597a5f55 | 142 | int i, cpu, me; |
6ef7a1bc RR |
143 | cpumask_var_t cpus; |
144 | bool called = true; | |
d9e368d6 | 145 | struct kvm_vcpu *vcpu; |
d9e368d6 | 146 | |
79f55997 | 147 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 148 | |
70e335e1 | 149 | raw_spin_lock(&kvm->requests_lock); |
e601e3be | 150 | me = smp_processor_id(); |
988a2cae | 151 | kvm_for_each_vcpu(i, vcpu, kvm) { |
a8eeb04a | 152 | if (kvm_make_check_request(req, vcpu)) |
d9e368d6 AK |
153 | continue; |
154 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
155 | if (cpus != NULL && cpu != -1 && cpu != me) |
156 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 157 | } |
6ef7a1bc RR |
158 | if (unlikely(cpus == NULL)) |
159 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
160 | else if (!cpumask_empty(cpus)) | |
161 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
162 | else | |
163 | called = false; | |
70e335e1 | 164 | raw_spin_unlock(&kvm->requests_lock); |
6ef7a1bc | 165 | free_cpumask_var(cpus); |
49846896 | 166 | return called; |
d9e368d6 AK |
167 | } |
168 | ||
49846896 | 169 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 170 | { |
49846896 RR |
171 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
172 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
173 | } |
174 | ||
49846896 RR |
175 | void kvm_reload_remote_mmus(struct kvm *kvm) |
176 | { | |
177 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
178 | } | |
2e53d63a | 179 | |
fb3f0f51 RR |
180 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
181 | { | |
182 | struct page *page; | |
183 | int r; | |
184 | ||
185 | mutex_init(&vcpu->mutex); | |
186 | vcpu->cpu = -1; | |
fb3f0f51 RR |
187 | vcpu->kvm = kvm; |
188 | vcpu->vcpu_id = id; | |
b6958ce4 | 189 | init_waitqueue_head(&vcpu->wq); |
af585b92 | 190 | kvm_async_pf_vcpu_init(vcpu); |
fb3f0f51 RR |
191 | |
192 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
193 | if (!page) { | |
194 | r = -ENOMEM; | |
195 | goto fail; | |
196 | } | |
197 | vcpu->run = page_address(page); | |
198 | ||
e9b11c17 | 199 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 200 | if (r < 0) |
e9b11c17 | 201 | goto fail_free_run; |
fb3f0f51 RR |
202 | return 0; |
203 | ||
fb3f0f51 RR |
204 | fail_free_run: |
205 | free_page((unsigned long)vcpu->run); | |
206 | fail: | |
76fafa5e | 207 | return r; |
fb3f0f51 RR |
208 | } |
209 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
210 | ||
211 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
212 | { | |
e9b11c17 | 213 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
214 | free_page((unsigned long)vcpu->run); |
215 | } | |
216 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
217 | ||
e930bffe AA |
218 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
219 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
220 | { | |
221 | return container_of(mn, struct kvm, mmu_notifier); | |
222 | } | |
223 | ||
224 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
225 | struct mm_struct *mm, | |
226 | unsigned long address) | |
227 | { | |
228 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 229 | int need_tlb_flush, idx; |
e930bffe AA |
230 | |
231 | /* | |
232 | * When ->invalidate_page runs, the linux pte has been zapped | |
233 | * already but the page is still allocated until | |
234 | * ->invalidate_page returns. So if we increase the sequence | |
235 | * here the kvm page fault will notice if the spte can't be | |
236 | * established because the page is going to be freed. If | |
237 | * instead the kvm page fault establishes the spte before | |
238 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
239 | * before returning. | |
240 | * | |
241 | * The sequence increase only need to be seen at spin_unlock | |
242 | * time, and not at spin_lock time. | |
243 | * | |
244 | * Increasing the sequence after the spin_unlock would be | |
245 | * unsafe because the kvm page fault could then establish the | |
246 | * pte after kvm_unmap_hva returned, without noticing the page | |
247 | * is going to be freed. | |
248 | */ | |
bc6678a3 | 249 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
250 | spin_lock(&kvm->mmu_lock); |
251 | kvm->mmu_notifier_seq++; | |
252 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
253 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 254 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
255 | |
256 | /* we've to flush the tlb before the pages can be freed */ | |
257 | if (need_tlb_flush) | |
258 | kvm_flush_remote_tlbs(kvm); | |
259 | ||
260 | } | |
261 | ||
3da0dd43 IE |
262 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
263 | struct mm_struct *mm, | |
264 | unsigned long address, | |
265 | pte_t pte) | |
266 | { | |
267 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 268 | int idx; |
3da0dd43 | 269 | |
bc6678a3 | 270 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
271 | spin_lock(&kvm->mmu_lock); |
272 | kvm->mmu_notifier_seq++; | |
273 | kvm_set_spte_hva(kvm, address, pte); | |
274 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 275 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
276 | } |
277 | ||
e930bffe AA |
278 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
279 | struct mm_struct *mm, | |
280 | unsigned long start, | |
281 | unsigned long end) | |
282 | { | |
283 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 284 | int need_tlb_flush = 0, idx; |
e930bffe | 285 | |
bc6678a3 | 286 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
287 | spin_lock(&kvm->mmu_lock); |
288 | /* | |
289 | * The count increase must become visible at unlock time as no | |
290 | * spte can be established without taking the mmu_lock and | |
291 | * count is also read inside the mmu_lock critical section. | |
292 | */ | |
293 | kvm->mmu_notifier_count++; | |
294 | for (; start < end; start += PAGE_SIZE) | |
295 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
296 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 297 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
298 | |
299 | /* we've to flush the tlb before the pages can be freed */ | |
300 | if (need_tlb_flush) | |
301 | kvm_flush_remote_tlbs(kvm); | |
302 | } | |
303 | ||
304 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
305 | struct mm_struct *mm, | |
306 | unsigned long start, | |
307 | unsigned long end) | |
308 | { | |
309 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
310 | ||
311 | spin_lock(&kvm->mmu_lock); | |
312 | /* | |
313 | * This sequence increase will notify the kvm page fault that | |
314 | * the page that is going to be mapped in the spte could have | |
315 | * been freed. | |
316 | */ | |
317 | kvm->mmu_notifier_seq++; | |
318 | /* | |
319 | * The above sequence increase must be visible before the | |
320 | * below count decrease but both values are read by the kvm | |
321 | * page fault under mmu_lock spinlock so we don't need to add | |
322 | * a smb_wmb() here in between the two. | |
323 | */ | |
324 | kvm->mmu_notifier_count--; | |
325 | spin_unlock(&kvm->mmu_lock); | |
326 | ||
327 | BUG_ON(kvm->mmu_notifier_count < 0); | |
328 | } | |
329 | ||
330 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
331 | struct mm_struct *mm, | |
332 | unsigned long address) | |
333 | { | |
334 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 335 | int young, idx; |
e930bffe | 336 | |
bc6678a3 | 337 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
338 | spin_lock(&kvm->mmu_lock); |
339 | young = kvm_age_hva(kvm, address); | |
340 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 341 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
342 | |
343 | if (young) | |
344 | kvm_flush_remote_tlbs(kvm); | |
345 | ||
346 | return young; | |
347 | } | |
348 | ||
85db06e5 MT |
349 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
350 | struct mm_struct *mm) | |
351 | { | |
352 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
353 | int idx; |
354 | ||
355 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 356 | kvm_arch_flush_shadow(kvm); |
eda2beda | 357 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
358 | } |
359 | ||
e930bffe AA |
360 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
361 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
362 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
363 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
364 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
3da0dd43 | 365 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 366 | .release = kvm_mmu_notifier_release, |
e930bffe | 367 | }; |
4c07b0a4 AK |
368 | |
369 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
370 | { | |
371 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
372 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
373 | } | |
374 | ||
375 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
376 | ||
377 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
378 | { | |
379 | return 0; | |
380 | } | |
381 | ||
e930bffe AA |
382 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
383 | ||
f17abe9a | 384 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 385 | { |
e93f8a0f | 386 | int r = 0, i; |
d19a9cd2 | 387 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 388 | |
d19a9cd2 ZX |
389 | if (IS_ERR(kvm)) |
390 | goto out; | |
10474ae8 AG |
391 | |
392 | r = hardware_enable_all(); | |
393 | if (r) | |
394 | goto out_err_nodisable; | |
395 | ||
75858a84 AK |
396 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
397 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 398 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 399 | #endif |
6aa8b732 | 400 | |
46a26bf5 MT |
401 | r = -ENOMEM; |
402 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
403 | if (!kvm->memslots) | |
404 | goto out_err; | |
bc6678a3 MT |
405 | if (init_srcu_struct(&kvm->srcu)) |
406 | goto out_err; | |
e93f8a0f MT |
407 | for (i = 0; i < KVM_NR_BUSES; i++) { |
408 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
409 | GFP_KERNEL); | |
410 | if (!kvm->buses[i]) { | |
411 | cleanup_srcu_struct(&kvm->srcu); | |
412 | goto out_err; | |
413 | } | |
414 | } | |
46a26bf5 | 415 | |
4c07b0a4 | 416 | r = kvm_init_mmu_notifier(kvm); |
283d0c65 | 417 | if (r) { |
bc6678a3 | 418 | cleanup_srcu_struct(&kvm->srcu); |
283d0c65 | 419 | goto out_err; |
e930bffe | 420 | } |
e930bffe | 421 | |
6d4e4c4f AK |
422 | kvm->mm = current->mm; |
423 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 424 | spin_lock_init(&kvm->mmu_lock); |
70e335e1 | 425 | raw_spin_lock_init(&kvm->requests_lock); |
d34e6b17 | 426 | kvm_eventfd_init(kvm); |
11ec2804 | 427 | mutex_init(&kvm->lock); |
60eead79 | 428 | mutex_init(&kvm->irq_lock); |
79fac95e | 429 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 430 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
431 | spin_lock(&kvm_lock); |
432 | list_add(&kvm->vm_list, &vm_list); | |
433 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 434 | out: |
f17abe9a | 435 | return kvm; |
10474ae8 AG |
436 | |
437 | out_err: | |
438 | hardware_disable_all(); | |
439 | out_err_nodisable: | |
e93f8a0f MT |
440 | for (i = 0; i < KVM_NR_BUSES; i++) |
441 | kfree(kvm->buses[i]); | |
46a26bf5 | 442 | kfree(kvm->memslots); |
10474ae8 AG |
443 | kfree(kvm); |
444 | return ERR_PTR(r); | |
f17abe9a AK |
445 | } |
446 | ||
6aa8b732 AK |
447 | /* |
448 | * Free any memory in @free but not in @dont. | |
449 | */ | |
450 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
451 | struct kvm_memory_slot *dont) | |
452 | { | |
ec04b260 JR |
453 | int i; |
454 | ||
290fc38d IE |
455 | if (!dont || free->rmap != dont->rmap) |
456 | vfree(free->rmap); | |
6aa8b732 AK |
457 | |
458 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
459 | vfree(free->dirty_bitmap); | |
460 | ||
ec04b260 JR |
461 | |
462 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
463 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
464 | vfree(free->lpage_info[i]); | |
465 | free->lpage_info[i] = NULL; | |
466 | } | |
467 | } | |
05da4558 | 468 | |
6aa8b732 | 469 | free->npages = 0; |
8b6d44c7 | 470 | free->dirty_bitmap = NULL; |
8d4e1288 | 471 | free->rmap = NULL; |
6aa8b732 AK |
472 | } |
473 | ||
d19a9cd2 | 474 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
475 | { |
476 | int i; | |
46a26bf5 MT |
477 | struct kvm_memslots *slots = kvm->memslots; |
478 | ||
479 | for (i = 0; i < slots->nmemslots; ++i) | |
480 | kvm_free_physmem_slot(&slots->memslots[i], NULL); | |
6aa8b732 | 481 | |
46a26bf5 | 482 | kfree(kvm->memslots); |
6aa8b732 AK |
483 | } |
484 | ||
f17abe9a AK |
485 | static void kvm_destroy_vm(struct kvm *kvm) |
486 | { | |
e93f8a0f | 487 | int i; |
6d4e4c4f AK |
488 | struct mm_struct *mm = kvm->mm; |
489 | ||
ad8ba2cd | 490 | kvm_arch_sync_events(kvm); |
133de902 AK |
491 | spin_lock(&kvm_lock); |
492 | list_del(&kvm->vm_list); | |
493 | spin_unlock(&kvm_lock); | |
399ec807 | 494 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
495 | for (i = 0; i < KVM_NR_BUSES; i++) |
496 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 497 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
498 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
499 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
500 | #else |
501 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 502 | #endif |
d19a9cd2 | 503 | kvm_arch_destroy_vm(kvm); |
10474ae8 | 504 | hardware_disable_all(); |
6d4e4c4f | 505 | mmdrop(mm); |
f17abe9a AK |
506 | } |
507 | ||
d39f13b0 IE |
508 | void kvm_get_kvm(struct kvm *kvm) |
509 | { | |
510 | atomic_inc(&kvm->users_count); | |
511 | } | |
512 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
513 | ||
514 | void kvm_put_kvm(struct kvm *kvm) | |
515 | { | |
516 | if (atomic_dec_and_test(&kvm->users_count)) | |
517 | kvm_destroy_vm(kvm); | |
518 | } | |
519 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
520 | ||
521 | ||
f17abe9a AK |
522 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
523 | { | |
524 | struct kvm *kvm = filp->private_data; | |
525 | ||
721eecbf GH |
526 | kvm_irqfd_release(kvm); |
527 | ||
d39f13b0 | 528 | kvm_put_kvm(kvm); |
6aa8b732 AK |
529 | return 0; |
530 | } | |
531 | ||
6aa8b732 AK |
532 | /* |
533 | * Allocate some memory and give it an address in the guest physical address | |
534 | * space. | |
535 | * | |
536 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 537 | * |
10589a46 | 538 | * Must be called holding mmap_sem for write. |
6aa8b732 | 539 | */ |
f78e0e2e SY |
540 | int __kvm_set_memory_region(struct kvm *kvm, |
541 | struct kvm_userspace_memory_region *mem, | |
542 | int user_alloc) | |
6aa8b732 | 543 | { |
bc6678a3 | 544 | int r, flush_shadow = 0; |
6aa8b732 | 545 | gfn_t base_gfn; |
28bcb112 HC |
546 | unsigned long npages; |
547 | unsigned long i; | |
6aa8b732 AK |
548 | struct kvm_memory_slot *memslot; |
549 | struct kvm_memory_slot old, new; | |
bc6678a3 | 550 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
551 | |
552 | r = -EINVAL; | |
553 | /* General sanity checks */ | |
554 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
555 | goto out; | |
556 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
557 | goto out; | |
e7cacd40 | 558 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 559 | goto out; |
e0d62c7f | 560 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
561 | goto out; |
562 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
563 | goto out; | |
564 | ||
46a26bf5 | 565 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
566 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
567 | npages = mem->memory_size >> PAGE_SHIFT; | |
568 | ||
660c22c4 TY |
569 | r = -EINVAL; |
570 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
571 | goto out; | |
572 | ||
6aa8b732 AK |
573 | if (!npages) |
574 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
575 | ||
6aa8b732 AK |
576 | new = old = *memslot; |
577 | ||
e36d96f7 | 578 | new.id = mem->slot; |
6aa8b732 AK |
579 | new.base_gfn = base_gfn; |
580 | new.npages = npages; | |
581 | new.flags = mem->flags; | |
582 | ||
583 | /* Disallow changing a memory slot's size. */ | |
584 | r = -EINVAL; | |
585 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 586 | goto out_free; |
6aa8b732 AK |
587 | |
588 | /* Check for overlaps */ | |
589 | r = -EEXIST; | |
590 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 591 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 592 | |
4cd481f6 | 593 | if (s == memslot || !s->npages) |
6aa8b732 AK |
594 | continue; |
595 | if (!((base_gfn + npages <= s->base_gfn) || | |
596 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 597 | goto out_free; |
6aa8b732 | 598 | } |
6aa8b732 | 599 | |
6aa8b732 AK |
600 | /* Free page dirty bitmap if unneeded */ |
601 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 602 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
603 | |
604 | r = -ENOMEM; | |
605 | ||
606 | /* Allocate if a slot is being created */ | |
eff0114a | 607 | #ifndef CONFIG_S390 |
8d4e1288 | 608 | if (npages && !new.rmap) { |
3bd89007 | 609 | new.rmap = vmalloc(npages * sizeof(*new.rmap)); |
290fc38d IE |
610 | |
611 | if (!new.rmap) | |
f78e0e2e | 612 | goto out_free; |
290fc38d | 613 | |
290fc38d | 614 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 615 | |
80b14b5b | 616 | new.user_alloc = user_alloc; |
bc6678a3 | 617 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 618 | } |
ec04b260 JR |
619 | if (!npages) |
620 | goto skip_lpage; | |
05da4558 | 621 | |
ec04b260 | 622 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
623 | unsigned long ugfn; |
624 | unsigned long j; | |
625 | int lpages; | |
ec04b260 | 626 | int level = i + 2; |
05da4558 | 627 | |
ec04b260 JR |
628 | /* Avoid unused variable warning if no large pages */ |
629 | (void)level; | |
630 | ||
631 | if (new.lpage_info[i]) | |
632 | continue; | |
633 | ||
82855413 JR |
634 | lpages = 1 + ((base_gfn + npages - 1) |
635 | >> KVM_HPAGE_GFN_SHIFT(level)); | |
636 | lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level); | |
ec04b260 JR |
637 | |
638 | new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i])); | |
639 | ||
640 | if (!new.lpage_info[i]) | |
05da4558 MT |
641 | goto out_free; |
642 | ||
ec04b260 JR |
643 | memset(new.lpage_info[i], 0, |
644 | lpages * sizeof(*new.lpage_info[i])); | |
05da4558 | 645 | |
82855413 | 646 | if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 647 | new.lpage_info[i][0].write_count = 1; |
82855413 | 648 | if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 649 | new.lpage_info[i][lpages - 1].write_count = 1; |
ac04527f AK |
650 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
651 | /* | |
652 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
653 | * other, or if explicitly asked to, disable large page |
654 | * support for this slot | |
ac04527f | 655 | */ |
ec04b260 | 656 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 657 | !largepages_enabled) |
ec04b260 JR |
658 | for (j = 0; j < lpages; ++j) |
659 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 660 | } |
6aa8b732 | 661 | |
ec04b260 JR |
662 | skip_lpage: |
663 | ||
6aa8b732 AK |
664 | /* Allocate page dirty bitmap if needed */ |
665 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
87bf6e7d | 666 | unsigned long dirty_bytes = kvm_dirty_bitmap_bytes(&new); |
6aa8b732 AK |
667 | |
668 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
669 | if (!new.dirty_bitmap) | |
f78e0e2e | 670 | goto out_free; |
6aa8b732 | 671 | memset(new.dirty_bitmap, 0, dirty_bytes); |
bc6678a3 | 672 | /* destroy any largepage mappings for dirty tracking */ |
e244584f | 673 | if (old.npages) |
bc6678a3 | 674 | flush_shadow = 1; |
6aa8b732 | 675 | } |
3eea8437 CB |
676 | #else /* not defined CONFIG_S390 */ |
677 | new.user_alloc = user_alloc; | |
678 | if (user_alloc) | |
679 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 680 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 681 | |
bc6678a3 MT |
682 | if (!npages) { |
683 | r = -ENOMEM; | |
684 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
685 | if (!slots) | |
686 | goto out_free; | |
687 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
688 | if (mem->slot >= slots->nmemslots) | |
689 | slots->nmemslots = mem->slot + 1; | |
690 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; | |
691 | ||
692 | old_memslots = kvm->memslots; | |
693 | rcu_assign_pointer(kvm->memslots, slots); | |
694 | synchronize_srcu_expedited(&kvm->srcu); | |
695 | /* From this point no new shadow pages pointing to a deleted | |
696 | * memslot will be created. | |
697 | * | |
698 | * validation of sp->gfn happens in: | |
699 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
700 | * - kvm_is_visible_gfn (mmu_check_roots) | |
701 | */ | |
34d4cb8f | 702 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
703 | kfree(old_memslots); |
704 | } | |
34d4cb8f | 705 | |
f7784b8e MT |
706 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
707 | if (r) | |
708 | goto out_free; | |
709 | ||
bc6678a3 MT |
710 | /* map the pages in iommu page table */ |
711 | if (npages) { | |
712 | r = kvm_iommu_map_pages(kvm, &new); | |
713 | if (r) | |
714 | goto out_free; | |
715 | } | |
604b38ac | 716 | |
bc6678a3 MT |
717 | r = -ENOMEM; |
718 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
719 | if (!slots) | |
720 | goto out_free; | |
721 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
722 | if (mem->slot >= slots->nmemslots) | |
723 | slots->nmemslots = mem->slot + 1; | |
724 | ||
725 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
726 | if (!npages) { | |
727 | new.rmap = NULL; | |
728 | new.dirty_bitmap = NULL; | |
729 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
730 | new.lpage_info[i] = NULL; | |
731 | } | |
732 | ||
733 | slots->memslots[mem->slot] = new; | |
734 | old_memslots = kvm->memslots; | |
735 | rcu_assign_pointer(kvm->memslots, slots); | |
736 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 737 | |
f7784b8e | 738 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 739 | |
bc6678a3 MT |
740 | kvm_free_physmem_slot(&old, &new); |
741 | kfree(old_memslots); | |
742 | ||
743 | if (flush_shadow) | |
744 | kvm_arch_flush_shadow(kvm); | |
745 | ||
6aa8b732 AK |
746 | return 0; |
747 | ||
f78e0e2e | 748 | out_free: |
6aa8b732 AK |
749 | kvm_free_physmem_slot(&new, &old); |
750 | out: | |
751 | return r; | |
210c7c4d IE |
752 | |
753 | } | |
f78e0e2e SY |
754 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
755 | ||
756 | int kvm_set_memory_region(struct kvm *kvm, | |
757 | struct kvm_userspace_memory_region *mem, | |
758 | int user_alloc) | |
759 | { | |
760 | int r; | |
761 | ||
79fac95e | 762 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 763 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 764 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
765 | return r; |
766 | } | |
210c7c4d IE |
767 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
768 | ||
1fe779f8 CO |
769 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
770 | struct | |
771 | kvm_userspace_memory_region *mem, | |
772 | int user_alloc) | |
210c7c4d | 773 | { |
e0d62c7f IE |
774 | if (mem->slot >= KVM_MEMORY_SLOTS) |
775 | return -EINVAL; | |
210c7c4d | 776 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
777 | } |
778 | ||
5bb064dc ZX |
779 | int kvm_get_dirty_log(struct kvm *kvm, |
780 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
781 | { |
782 | struct kvm_memory_slot *memslot; | |
783 | int r, i; | |
87bf6e7d | 784 | unsigned long n; |
6aa8b732 AK |
785 | unsigned long any = 0; |
786 | ||
6aa8b732 AK |
787 | r = -EINVAL; |
788 | if (log->slot >= KVM_MEMORY_SLOTS) | |
789 | goto out; | |
790 | ||
46a26bf5 | 791 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
792 | r = -ENOENT; |
793 | if (!memslot->dirty_bitmap) | |
794 | goto out; | |
795 | ||
87bf6e7d | 796 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 797 | |
cd1a4a98 | 798 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
799 | any = memslot->dirty_bitmap[i]; |
800 | ||
801 | r = -EFAULT; | |
802 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
803 | goto out; | |
804 | ||
5bb064dc ZX |
805 | if (any) |
806 | *is_dirty = 1; | |
6aa8b732 AK |
807 | |
808 | r = 0; | |
6aa8b732 | 809 | out: |
6aa8b732 AK |
810 | return r; |
811 | } | |
812 | ||
54dee993 MT |
813 | void kvm_disable_largepages(void) |
814 | { | |
815 | largepages_enabled = false; | |
816 | } | |
817 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
818 | ||
cea7bb21 IE |
819 | int is_error_page(struct page *page) |
820 | { | |
edba23e5 | 821 | return page == bad_page || page == hwpoison_page || page == fault_page; |
cea7bb21 IE |
822 | } |
823 | EXPORT_SYMBOL_GPL(is_error_page); | |
824 | ||
35149e21 AL |
825 | int is_error_pfn(pfn_t pfn) |
826 | { | |
edba23e5 | 827 | return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn; |
35149e21 AL |
828 | } |
829 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
830 | ||
bf998156 HY |
831 | int is_hwpoison_pfn(pfn_t pfn) |
832 | { | |
833 | return pfn == hwpoison_pfn; | |
834 | } | |
835 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
836 | ||
edba23e5 GN |
837 | int is_fault_pfn(pfn_t pfn) |
838 | { | |
839 | return pfn == fault_pfn; | |
840 | } | |
841 | EXPORT_SYMBOL_GPL(is_fault_pfn); | |
842 | ||
f9d46eb0 IE |
843 | static inline unsigned long bad_hva(void) |
844 | { | |
845 | return PAGE_OFFSET; | |
846 | } | |
847 | ||
848 | int kvm_is_error_hva(unsigned long addr) | |
849 | { | |
850 | return addr == bad_hva(); | |
851 | } | |
852 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
853 | ||
a1f4d395 | 854 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
855 | { |
856 | int i; | |
90d83dc3 | 857 | struct kvm_memslots *slots = kvm_memslots(kvm); |
6aa8b732 | 858 | |
46a26bf5 MT |
859 | for (i = 0; i < slots->nmemslots; ++i) { |
860 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
861 | |
862 | if (gfn >= memslot->base_gfn | |
863 | && gfn < memslot->base_gfn + memslot->npages) | |
864 | return memslot; | |
865 | } | |
8b6d44c7 | 866 | return NULL; |
6aa8b732 | 867 | } |
a1f4d395 | 868 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 869 | |
e0d62c7f IE |
870 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
871 | { | |
872 | int i; | |
90d83dc3 | 873 | struct kvm_memslots *slots = kvm_memslots(kvm); |
e0d62c7f | 874 | |
e0d62c7f | 875 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 876 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 877 | |
bc6678a3 MT |
878 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
879 | continue; | |
880 | ||
e0d62c7f IE |
881 | if (gfn >= memslot->base_gfn |
882 | && gfn < memslot->base_gfn + memslot->npages) | |
883 | return 1; | |
884 | } | |
885 | return 0; | |
886 | } | |
887 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
888 | ||
8f0b1ab6 JR |
889 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
890 | { | |
891 | struct vm_area_struct *vma; | |
892 | unsigned long addr, size; | |
893 | ||
894 | size = PAGE_SIZE; | |
895 | ||
896 | addr = gfn_to_hva(kvm, gfn); | |
897 | if (kvm_is_error_hva(addr)) | |
898 | return PAGE_SIZE; | |
899 | ||
900 | down_read(¤t->mm->mmap_sem); | |
901 | vma = find_vma(current->mm, addr); | |
902 | if (!vma) | |
903 | goto out; | |
904 | ||
905 | size = vma_kernel_pagesize(vma); | |
906 | ||
907 | out: | |
908 | up_read(¤t->mm->mmap_sem); | |
909 | ||
910 | return size; | |
911 | } | |
912 | ||
bc6678a3 MT |
913 | int memslot_id(struct kvm *kvm, gfn_t gfn) |
914 | { | |
915 | int i; | |
90d83dc3 | 916 | struct kvm_memslots *slots = kvm_memslots(kvm); |
bc6678a3 MT |
917 | struct kvm_memory_slot *memslot = NULL; |
918 | ||
bc6678a3 MT |
919 | for (i = 0; i < slots->nmemslots; ++i) { |
920 | memslot = &slots->memslots[i]; | |
921 | ||
922 | if (gfn >= memslot->base_gfn | |
923 | && gfn < memslot->base_gfn + memslot->npages) | |
924 | break; | |
925 | } | |
926 | ||
927 | return memslot - slots->memslots; | |
928 | } | |
929 | ||
48987781 XG |
930 | static unsigned long gfn_to_hva_many(struct kvm *kvm, gfn_t gfn, |
931 | gfn_t *nr_pages) | |
539cb660 IE |
932 | { |
933 | struct kvm_memory_slot *slot; | |
934 | ||
a1f4d395 | 935 | slot = gfn_to_memslot(kvm, gfn); |
bc6678a3 | 936 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 937 | return bad_hva(); |
48987781 XG |
938 | |
939 | if (nr_pages) | |
940 | *nr_pages = slot->npages - (gfn - slot->base_gfn); | |
941 | ||
f5c98031 | 942 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 943 | } |
48987781 XG |
944 | |
945 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) | |
946 | { | |
947 | return gfn_to_hva_many(kvm, gfn, NULL); | |
948 | } | |
0d150298 | 949 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 950 | |
af585b92 GN |
951 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic, |
952 | bool *async) | |
954bbbc2 | 953 | { |
8d4e1288 | 954 | struct page *page[1]; |
af585b92 | 955 | int npages = 0; |
2e2e3738 | 956 | pfn_t pfn; |
954bbbc2 | 957 | |
af585b92 GN |
958 | /* we can do it either atomically or asynchronously, not both */ |
959 | BUG_ON(atomic && async); | |
960 | ||
961 | if (atomic || async) | |
887c08ac | 962 | npages = __get_user_pages_fast(addr, 1, 1, page); |
af585b92 GN |
963 | |
964 | if (unlikely(npages != 1) && !atomic) { | |
887c08ac XG |
965 | might_sleep(); |
966 | npages = get_user_pages_fast(addr, 1, 1, page); | |
967 | } | |
539cb660 | 968 | |
2e2e3738 AL |
969 | if (unlikely(npages != 1)) { |
970 | struct vm_area_struct *vma; | |
971 | ||
887c08ac XG |
972 | if (atomic) |
973 | goto return_fault_page; | |
974 | ||
bbeb3406 | 975 | down_read(¤t->mm->mmap_sem); |
bf998156 | 976 | if (is_hwpoison_address(addr)) { |
bbeb3406 | 977 | up_read(¤t->mm->mmap_sem); |
bf998156 HY |
978 | get_page(hwpoison_page); |
979 | return page_to_pfn(hwpoison_page); | |
980 | } | |
981 | ||
2e2e3738 | 982 | vma = find_vma(current->mm, addr); |
4c2155ce | 983 | |
2e2e3738 AL |
984 | if (vma == NULL || addr < vma->vm_start || |
985 | !(vma->vm_flags & VM_PFNMAP)) { | |
af585b92 GN |
986 | if (async && !(vma->vm_flags & VM_PFNMAP) && |
987 | (vma->vm_flags & VM_WRITE)) | |
988 | *async = true; | |
4c2155ce | 989 | up_read(¤t->mm->mmap_sem); |
887c08ac | 990 | return_fault_page: |
edba23e5 GN |
991 | get_page(fault_page); |
992 | return page_to_pfn(fault_page); | |
2e2e3738 AL |
993 | } |
994 | ||
995 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 996 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 997 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
998 | } else |
999 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1000 | |
2e2e3738 | 1001 | return pfn; |
35149e21 AL |
1002 | } |
1003 | ||
887c08ac XG |
1004 | pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr) |
1005 | { | |
af585b92 | 1006 | return hva_to_pfn(kvm, addr, true, NULL); |
887c08ac XG |
1007 | } |
1008 | EXPORT_SYMBOL_GPL(hva_to_pfn_atomic); | |
1009 | ||
af585b92 | 1010 | static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async) |
506f0d6f MT |
1011 | { |
1012 | unsigned long addr; | |
1013 | ||
af585b92 GN |
1014 | if (async) |
1015 | *async = false; | |
1016 | ||
506f0d6f MT |
1017 | addr = gfn_to_hva(kvm, gfn); |
1018 | if (kvm_is_error_hva(addr)) { | |
1019 | get_page(bad_page); | |
1020 | return page_to_pfn(bad_page); | |
1021 | } | |
1022 | ||
af585b92 | 1023 | return hva_to_pfn(kvm, addr, atomic, async); |
365fb3fd XG |
1024 | } |
1025 | ||
1026 | pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn) | |
1027 | { | |
af585b92 | 1028 | return __gfn_to_pfn(kvm, gfn, true, NULL); |
365fb3fd XG |
1029 | } |
1030 | EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic); | |
1031 | ||
af585b92 GN |
1032 | pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async) |
1033 | { | |
1034 | return __gfn_to_pfn(kvm, gfn, false, async); | |
1035 | } | |
1036 | EXPORT_SYMBOL_GPL(gfn_to_pfn_async); | |
1037 | ||
365fb3fd XG |
1038 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
1039 | { | |
af585b92 | 1040 | return __gfn_to_pfn(kvm, gfn, false, NULL); |
506f0d6f | 1041 | } |
35149e21 AL |
1042 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
1043 | ||
506f0d6f MT |
1044 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
1045 | struct kvm_memory_slot *slot, gfn_t gfn) | |
1046 | { | |
1047 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
af585b92 | 1048 | return hva_to_pfn(kvm, addr, false, NULL); |
506f0d6f MT |
1049 | } |
1050 | ||
48987781 XG |
1051 | int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages, |
1052 | int nr_pages) | |
1053 | { | |
1054 | unsigned long addr; | |
1055 | gfn_t entry; | |
1056 | ||
1057 | addr = gfn_to_hva_many(kvm, gfn, &entry); | |
1058 | if (kvm_is_error_hva(addr)) | |
1059 | return -1; | |
1060 | ||
1061 | if (entry < nr_pages) | |
1062 | return 0; | |
1063 | ||
1064 | return __get_user_pages_fast(addr, nr_pages, 1, pages); | |
1065 | } | |
1066 | EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic); | |
1067 | ||
35149e21 AL |
1068 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1069 | { | |
2e2e3738 AL |
1070 | pfn_t pfn; |
1071 | ||
1072 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1073 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1074 | return pfn_to_page(pfn); |
1075 | ||
c77fb9dc | 1076 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1077 | |
1078 | get_page(bad_page); | |
1079 | return bad_page; | |
954bbbc2 | 1080 | } |
aab61cc0 | 1081 | |
954bbbc2 AK |
1082 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1083 | ||
b4231d61 IE |
1084 | void kvm_release_page_clean(struct page *page) |
1085 | { | |
35149e21 | 1086 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1087 | } |
1088 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1089 | ||
35149e21 AL |
1090 | void kvm_release_pfn_clean(pfn_t pfn) |
1091 | { | |
c77fb9dc | 1092 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1093 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1094 | } |
1095 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1096 | ||
b4231d61 | 1097 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1098 | { |
35149e21 AL |
1099 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1100 | } | |
1101 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1102 | ||
1103 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1104 | { | |
1105 | kvm_set_pfn_dirty(pfn); | |
1106 | kvm_release_pfn_clean(pfn); | |
1107 | } | |
1108 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1109 | ||
1110 | void kvm_set_page_dirty(struct page *page) | |
1111 | { | |
1112 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1113 | } | |
1114 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1115 | ||
1116 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1117 | { | |
c77fb9dc | 1118 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1119 | struct page *page = pfn_to_page(pfn); |
1120 | if (!PageReserved(page)) | |
1121 | SetPageDirty(page); | |
1122 | } | |
8a7ae055 | 1123 | } |
35149e21 AL |
1124 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1125 | ||
1126 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1127 | { | |
c77fb9dc | 1128 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1129 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1130 | } |
1131 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1132 | ||
1133 | void kvm_get_pfn(pfn_t pfn) | |
1134 | { | |
c77fb9dc | 1135 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1136 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1137 | } |
1138 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1139 | |
195aefde IE |
1140 | static int next_segment(unsigned long len, int offset) |
1141 | { | |
1142 | if (len > PAGE_SIZE - offset) | |
1143 | return PAGE_SIZE - offset; | |
1144 | else | |
1145 | return len; | |
1146 | } | |
1147 | ||
1148 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1149 | int len) | |
1150 | { | |
e0506bcb IE |
1151 | int r; |
1152 | unsigned long addr; | |
195aefde | 1153 | |
e0506bcb IE |
1154 | addr = gfn_to_hva(kvm, gfn); |
1155 | if (kvm_is_error_hva(addr)) | |
1156 | return -EFAULT; | |
1157 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1158 | if (r) | |
195aefde | 1159 | return -EFAULT; |
195aefde IE |
1160 | return 0; |
1161 | } | |
1162 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1163 | ||
1164 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1165 | { | |
1166 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1167 | int seg; | |
1168 | int offset = offset_in_page(gpa); | |
1169 | int ret; | |
1170 | ||
1171 | while ((seg = next_segment(len, offset)) != 0) { | |
1172 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1173 | if (ret < 0) | |
1174 | return ret; | |
1175 | offset = 0; | |
1176 | len -= seg; | |
1177 | data += seg; | |
1178 | ++gfn; | |
1179 | } | |
1180 | return 0; | |
1181 | } | |
1182 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1183 | ||
7ec54588 MT |
1184 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1185 | unsigned long len) | |
1186 | { | |
1187 | int r; | |
1188 | unsigned long addr; | |
1189 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1190 | int offset = offset_in_page(gpa); | |
1191 | ||
1192 | addr = gfn_to_hva(kvm, gfn); | |
1193 | if (kvm_is_error_hva(addr)) | |
1194 | return -EFAULT; | |
0aac03f0 | 1195 | pagefault_disable(); |
7ec54588 | 1196 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1197 | pagefault_enable(); |
7ec54588 MT |
1198 | if (r) |
1199 | return -EFAULT; | |
1200 | return 0; | |
1201 | } | |
1202 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1203 | ||
195aefde IE |
1204 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1205 | int offset, int len) | |
1206 | { | |
e0506bcb IE |
1207 | int r; |
1208 | unsigned long addr; | |
195aefde | 1209 | |
e0506bcb IE |
1210 | addr = gfn_to_hva(kvm, gfn); |
1211 | if (kvm_is_error_hva(addr)) | |
1212 | return -EFAULT; | |
1213 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1214 | if (r) | |
195aefde | 1215 | return -EFAULT; |
195aefde IE |
1216 | mark_page_dirty(kvm, gfn); |
1217 | return 0; | |
1218 | } | |
1219 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1220 | ||
1221 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1222 | unsigned long len) | |
1223 | { | |
1224 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1225 | int seg; | |
1226 | int offset = offset_in_page(gpa); | |
1227 | int ret; | |
1228 | ||
1229 | while ((seg = next_segment(len, offset)) != 0) { | |
1230 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1231 | if (ret < 0) | |
1232 | return ret; | |
1233 | offset = 0; | |
1234 | len -= seg; | |
1235 | data += seg; | |
1236 | ++gfn; | |
1237 | } | |
1238 | return 0; | |
1239 | } | |
1240 | ||
1241 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1242 | { | |
3e021bf5 | 1243 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1244 | } |
1245 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1246 | ||
1247 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1248 | { | |
1249 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1250 | int seg; | |
1251 | int offset = offset_in_page(gpa); | |
1252 | int ret; | |
1253 | ||
1254 | while ((seg = next_segment(len, offset)) != 0) { | |
1255 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1256 | if (ret < 0) | |
1257 | return ret; | |
1258 | offset = 0; | |
1259 | len -= seg; | |
1260 | ++gfn; | |
1261 | } | |
1262 | return 0; | |
1263 | } | |
1264 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1265 | ||
6aa8b732 AK |
1266 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1267 | { | |
31389947 | 1268 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1269 | |
a1f4d395 | 1270 | memslot = gfn_to_memslot(kvm, gfn); |
7e9d619d RR |
1271 | if (memslot && memslot->dirty_bitmap) { |
1272 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1273 | |
d1476937 | 1274 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); |
6aa8b732 AK |
1275 | } |
1276 | } | |
1277 | ||
b6958ce4 ED |
1278 | /* |
1279 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1280 | */ | |
8776e519 | 1281 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1282 | { |
e5c239cf MT |
1283 | DEFINE_WAIT(wait); |
1284 | ||
1285 | for (;;) { | |
1286 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1287 | ||
a1b37100 | 1288 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1289 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1290 | break; |
d7690175 | 1291 | } |
09cec754 GN |
1292 | if (kvm_cpu_has_pending_timer(vcpu)) |
1293 | break; | |
e5c239cf MT |
1294 | if (signal_pending(current)) |
1295 | break; | |
1296 | ||
b6958ce4 | 1297 | schedule(); |
b6958ce4 | 1298 | } |
d3bef15f | 1299 | |
e5c239cf | 1300 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1301 | } |
1302 | ||
6aa8b732 AK |
1303 | void kvm_resched(struct kvm_vcpu *vcpu) |
1304 | { | |
3fca0365 YD |
1305 | if (!need_resched()) |
1306 | return; | |
6aa8b732 | 1307 | cond_resched(); |
6aa8b732 AK |
1308 | } |
1309 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1310 | ||
d255f4f2 ZE |
1311 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu) |
1312 | { | |
1313 | ktime_t expires; | |
1314 | DEFINE_WAIT(wait); | |
1315 | ||
1316 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1317 | ||
1318 | /* Sleep for 100 us, and hope lock-holder got scheduled */ | |
1319 | expires = ktime_add_ns(ktime_get(), 100000UL); | |
1320 | schedule_hrtimeout(&expires, HRTIMER_MODE_ABS); | |
1321 | ||
1322 | finish_wait(&vcpu->wq, &wait); | |
1323 | } | |
1324 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1325 | ||
e4a533a4 | 1326 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1327 | { |
1328 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1329 | struct page *page; |
1330 | ||
e4a533a4 | 1331 | if (vmf->pgoff == 0) |
039576c0 | 1332 | page = virt_to_page(vcpu->run); |
09566765 | 1333 | #ifdef CONFIG_X86 |
e4a533a4 | 1334 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1335 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1336 | #endif |
1337 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1338 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1339 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1340 | #endif |
039576c0 | 1341 | else |
e4a533a4 | 1342 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1343 | get_page(page); |
e4a533a4 | 1344 | vmf->page = page; |
1345 | return 0; | |
9a2bb7f4 AK |
1346 | } |
1347 | ||
f0f37e2f | 1348 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1349 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1350 | }; |
1351 | ||
1352 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1353 | { | |
1354 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1355 | return 0; | |
1356 | } | |
1357 | ||
bccf2150 AK |
1358 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1359 | { | |
1360 | struct kvm_vcpu *vcpu = filp->private_data; | |
1361 | ||
66c0b394 | 1362 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1363 | return 0; |
1364 | } | |
1365 | ||
3d3aab1b | 1366 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1367 | .release = kvm_vcpu_release, |
1368 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1369 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1370 | .mmap = kvm_vcpu_mmap, |
6038f373 | 1371 | .llseek = noop_llseek, |
bccf2150 AK |
1372 | }; |
1373 | ||
1374 | /* | |
1375 | * Allocates an inode for the vcpu. | |
1376 | */ | |
1377 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1378 | { | |
628ff7c1 | 1379 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1380 | } |
1381 | ||
c5ea7660 AK |
1382 | /* |
1383 | * Creates some virtual cpus. Good luck creating more than one. | |
1384 | */ | |
73880c80 | 1385 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1386 | { |
1387 | int r; | |
988a2cae | 1388 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1389 | |
73880c80 | 1390 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1391 | if (IS_ERR(vcpu)) |
1392 | return PTR_ERR(vcpu); | |
c5ea7660 | 1393 | |
15ad7146 AK |
1394 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1395 | ||
26e5215f AK |
1396 | r = kvm_arch_vcpu_setup(vcpu); |
1397 | if (r) | |
7d8fece6 | 1398 | return r; |
26e5215f | 1399 | |
11ec2804 | 1400 | mutex_lock(&kvm->lock); |
73880c80 GN |
1401 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1402 | r = -EINVAL; | |
e9b11c17 | 1403 | goto vcpu_destroy; |
fb3f0f51 | 1404 | } |
73880c80 | 1405 | |
988a2cae GN |
1406 | kvm_for_each_vcpu(r, v, kvm) |
1407 | if (v->vcpu_id == id) { | |
73880c80 GN |
1408 | r = -EEXIST; |
1409 | goto vcpu_destroy; | |
1410 | } | |
1411 | ||
1412 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1413 | |
fb3f0f51 | 1414 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1415 | kvm_get_kvm(kvm); |
bccf2150 | 1416 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1417 | if (r < 0) { |
1418 | kvm_put_kvm(kvm); | |
1419 | goto vcpu_destroy; | |
1420 | } | |
1421 | ||
1422 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1423 | smp_wmb(); | |
1424 | atomic_inc(&kvm->online_vcpus); | |
1425 | ||
1426 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1427 | if (kvm->bsp_vcpu_id == id) | |
1428 | kvm->bsp_vcpu = vcpu; | |
1429 | #endif | |
1430 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1431 | return r; |
39c3b86e | 1432 | |
e9b11c17 | 1433 | vcpu_destroy: |
7d8fece6 | 1434 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1435 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1436 | return r; |
1437 | } | |
1438 | ||
1961d276 AK |
1439 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1440 | { | |
1441 | if (sigset) { | |
1442 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1443 | vcpu->sigset_active = 1; | |
1444 | vcpu->sigset = *sigset; | |
1445 | } else | |
1446 | vcpu->sigset_active = 0; | |
1447 | return 0; | |
1448 | } | |
1449 | ||
bccf2150 AK |
1450 | static long kvm_vcpu_ioctl(struct file *filp, |
1451 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1452 | { |
bccf2150 | 1453 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1454 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1455 | int r; |
fa3795a7 DH |
1456 | struct kvm_fpu *fpu = NULL; |
1457 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1458 | |
6d4e4c4f AK |
1459 | if (vcpu->kvm->mm != current->mm) |
1460 | return -EIO; | |
2122ff5e AK |
1461 | |
1462 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1463 | /* | |
1464 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1465 | * so vcpu_load() would break it. | |
1466 | */ | |
1467 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1468 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1469 | #endif | |
1470 | ||
1471 | ||
1472 | vcpu_load(vcpu); | |
6aa8b732 | 1473 | switch (ioctl) { |
9a2bb7f4 | 1474 | case KVM_RUN: |
f0fe5108 AK |
1475 | r = -EINVAL; |
1476 | if (arg) | |
1477 | goto out; | |
b6c7a5dc | 1478 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1479 | break; |
6aa8b732 | 1480 | case KVM_GET_REGS: { |
3e4bb3ac | 1481 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1482 | |
3e4bb3ac XZ |
1483 | r = -ENOMEM; |
1484 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1485 | if (!kvm_regs) | |
6aa8b732 | 1486 | goto out; |
3e4bb3ac XZ |
1487 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1488 | if (r) | |
1489 | goto out_free1; | |
6aa8b732 | 1490 | r = -EFAULT; |
3e4bb3ac XZ |
1491 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1492 | goto out_free1; | |
6aa8b732 | 1493 | r = 0; |
3e4bb3ac XZ |
1494 | out_free1: |
1495 | kfree(kvm_regs); | |
6aa8b732 AK |
1496 | break; |
1497 | } | |
1498 | case KVM_SET_REGS: { | |
3e4bb3ac | 1499 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1500 | |
3e4bb3ac XZ |
1501 | r = -ENOMEM; |
1502 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1503 | if (!kvm_regs) | |
6aa8b732 | 1504 | goto out; |
3e4bb3ac XZ |
1505 | r = -EFAULT; |
1506 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1507 | goto out_free2; | |
1508 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1509 | if (r) |
3e4bb3ac | 1510 | goto out_free2; |
6aa8b732 | 1511 | r = 0; |
3e4bb3ac XZ |
1512 | out_free2: |
1513 | kfree(kvm_regs); | |
6aa8b732 AK |
1514 | break; |
1515 | } | |
1516 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1517 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1518 | r = -ENOMEM; | |
1519 | if (!kvm_sregs) | |
1520 | goto out; | |
1521 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1522 | if (r) |
1523 | goto out; | |
1524 | r = -EFAULT; | |
fa3795a7 | 1525 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1526 | goto out; |
1527 | r = 0; | |
1528 | break; | |
1529 | } | |
1530 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1531 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1532 | r = -ENOMEM; | |
1533 | if (!kvm_sregs) | |
1534 | goto out; | |
6aa8b732 | 1535 | r = -EFAULT; |
fa3795a7 | 1536 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1537 | goto out; |
fa3795a7 | 1538 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1539 | if (r) |
1540 | goto out; | |
1541 | r = 0; | |
1542 | break; | |
1543 | } | |
62d9f0db MT |
1544 | case KVM_GET_MP_STATE: { |
1545 | struct kvm_mp_state mp_state; | |
1546 | ||
1547 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1548 | if (r) | |
1549 | goto out; | |
1550 | r = -EFAULT; | |
1551 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1552 | goto out; | |
1553 | r = 0; | |
1554 | break; | |
1555 | } | |
1556 | case KVM_SET_MP_STATE: { | |
1557 | struct kvm_mp_state mp_state; | |
1558 | ||
1559 | r = -EFAULT; | |
1560 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1561 | goto out; | |
1562 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1563 | if (r) | |
1564 | goto out; | |
1565 | r = 0; | |
1566 | break; | |
1567 | } | |
6aa8b732 AK |
1568 | case KVM_TRANSLATE: { |
1569 | struct kvm_translation tr; | |
1570 | ||
1571 | r = -EFAULT; | |
2f366987 | 1572 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1573 | goto out; |
8b006791 | 1574 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1575 | if (r) |
1576 | goto out; | |
1577 | r = -EFAULT; | |
2f366987 | 1578 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1579 | goto out; |
1580 | r = 0; | |
1581 | break; | |
1582 | } | |
d0bfb940 JK |
1583 | case KVM_SET_GUEST_DEBUG: { |
1584 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1585 | |
1586 | r = -EFAULT; | |
2f366987 | 1587 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1588 | goto out; |
d0bfb940 | 1589 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1590 | if (r) |
1591 | goto out; | |
1592 | r = 0; | |
1593 | break; | |
1594 | } | |
1961d276 AK |
1595 | case KVM_SET_SIGNAL_MASK: { |
1596 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1597 | struct kvm_signal_mask kvm_sigmask; | |
1598 | sigset_t sigset, *p; | |
1599 | ||
1600 | p = NULL; | |
1601 | if (argp) { | |
1602 | r = -EFAULT; | |
1603 | if (copy_from_user(&kvm_sigmask, argp, | |
1604 | sizeof kvm_sigmask)) | |
1605 | goto out; | |
1606 | r = -EINVAL; | |
1607 | if (kvm_sigmask.len != sizeof sigset) | |
1608 | goto out; | |
1609 | r = -EFAULT; | |
1610 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1611 | sizeof sigset)) | |
1612 | goto out; | |
1613 | p = &sigset; | |
1614 | } | |
376d41ff | 1615 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1616 | break; |
1617 | } | |
b8836737 | 1618 | case KVM_GET_FPU: { |
fa3795a7 DH |
1619 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1620 | r = -ENOMEM; | |
1621 | if (!fpu) | |
1622 | goto out; | |
1623 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1624 | if (r) |
1625 | goto out; | |
1626 | r = -EFAULT; | |
fa3795a7 | 1627 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1628 | goto out; |
1629 | r = 0; | |
1630 | break; | |
1631 | } | |
1632 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1633 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1634 | r = -ENOMEM; | |
1635 | if (!fpu) | |
1636 | goto out; | |
b8836737 | 1637 | r = -EFAULT; |
fa3795a7 | 1638 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1639 | goto out; |
fa3795a7 | 1640 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1641 | if (r) |
1642 | goto out; | |
1643 | r = 0; | |
1644 | break; | |
1645 | } | |
bccf2150 | 1646 | default: |
313a3dc7 | 1647 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1648 | } |
1649 | out: | |
2122ff5e | 1650 | vcpu_put(vcpu); |
fa3795a7 DH |
1651 | kfree(fpu); |
1652 | kfree(kvm_sregs); | |
bccf2150 AK |
1653 | return r; |
1654 | } | |
1655 | ||
1656 | static long kvm_vm_ioctl(struct file *filp, | |
1657 | unsigned int ioctl, unsigned long arg) | |
1658 | { | |
1659 | struct kvm *kvm = filp->private_data; | |
1660 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1661 | int r; |
bccf2150 | 1662 | |
6d4e4c4f AK |
1663 | if (kvm->mm != current->mm) |
1664 | return -EIO; | |
bccf2150 AK |
1665 | switch (ioctl) { |
1666 | case KVM_CREATE_VCPU: | |
1667 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1668 | if (r < 0) | |
1669 | goto out; | |
1670 | break; | |
6fc138d2 IE |
1671 | case KVM_SET_USER_MEMORY_REGION: { |
1672 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1673 | ||
1674 | r = -EFAULT; | |
1675 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1676 | sizeof kvm_userspace_mem)) | |
1677 | goto out; | |
1678 | ||
1679 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1680 | if (r) |
1681 | goto out; | |
1682 | break; | |
1683 | } | |
1684 | case KVM_GET_DIRTY_LOG: { | |
1685 | struct kvm_dirty_log log; | |
1686 | ||
1687 | r = -EFAULT; | |
2f366987 | 1688 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1689 | goto out; |
2c6f5df9 | 1690 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1691 | if (r) |
1692 | goto out; | |
1693 | break; | |
1694 | } | |
5f94c174 LV |
1695 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1696 | case KVM_REGISTER_COALESCED_MMIO: { | |
1697 | struct kvm_coalesced_mmio_zone zone; | |
1698 | r = -EFAULT; | |
1699 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1700 | goto out; | |
5f94c174 LV |
1701 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
1702 | if (r) | |
1703 | goto out; | |
1704 | r = 0; | |
1705 | break; | |
1706 | } | |
1707 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1708 | struct kvm_coalesced_mmio_zone zone; | |
1709 | r = -EFAULT; | |
1710 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1711 | goto out; | |
5f94c174 LV |
1712 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
1713 | if (r) | |
1714 | goto out; | |
1715 | r = 0; | |
1716 | break; | |
1717 | } | |
1718 | #endif | |
721eecbf GH |
1719 | case KVM_IRQFD: { |
1720 | struct kvm_irqfd data; | |
1721 | ||
1722 | r = -EFAULT; | |
1723 | if (copy_from_user(&data, argp, sizeof data)) | |
1724 | goto out; | |
1725 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1726 | break; | |
1727 | } | |
d34e6b17 GH |
1728 | case KVM_IOEVENTFD: { |
1729 | struct kvm_ioeventfd data; | |
1730 | ||
1731 | r = -EFAULT; | |
1732 | if (copy_from_user(&data, argp, sizeof data)) | |
1733 | goto out; | |
1734 | r = kvm_ioeventfd(kvm, &data); | |
1735 | break; | |
1736 | } | |
73880c80 GN |
1737 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1738 | case KVM_SET_BOOT_CPU_ID: | |
1739 | r = 0; | |
894a9c55 | 1740 | mutex_lock(&kvm->lock); |
73880c80 GN |
1741 | if (atomic_read(&kvm->online_vcpus) != 0) |
1742 | r = -EBUSY; | |
1743 | else | |
1744 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1745 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1746 | break; |
1747 | #endif | |
f17abe9a | 1748 | default: |
1fe779f8 | 1749 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1750 | if (r == -ENOTTY) |
1751 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1752 | } |
1753 | out: | |
1754 | return r; | |
1755 | } | |
1756 | ||
6ff5894c AB |
1757 | #ifdef CONFIG_COMPAT |
1758 | struct compat_kvm_dirty_log { | |
1759 | __u32 slot; | |
1760 | __u32 padding1; | |
1761 | union { | |
1762 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1763 | __u64 padding2; | |
1764 | }; | |
1765 | }; | |
1766 | ||
1767 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1768 | unsigned int ioctl, unsigned long arg) | |
1769 | { | |
1770 | struct kvm *kvm = filp->private_data; | |
1771 | int r; | |
1772 | ||
1773 | if (kvm->mm != current->mm) | |
1774 | return -EIO; | |
1775 | switch (ioctl) { | |
1776 | case KVM_GET_DIRTY_LOG: { | |
1777 | struct compat_kvm_dirty_log compat_log; | |
1778 | struct kvm_dirty_log log; | |
1779 | ||
1780 | r = -EFAULT; | |
1781 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1782 | sizeof(compat_log))) | |
1783 | goto out; | |
1784 | log.slot = compat_log.slot; | |
1785 | log.padding1 = compat_log.padding1; | |
1786 | log.padding2 = compat_log.padding2; | |
1787 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
1788 | ||
1789 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
1790 | if (r) | |
1791 | goto out; | |
1792 | break; | |
1793 | } | |
1794 | default: | |
1795 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
1796 | } | |
1797 | ||
1798 | out: | |
1799 | return r; | |
1800 | } | |
1801 | #endif | |
1802 | ||
e4a533a4 | 1803 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1804 | { |
777b3f49 MT |
1805 | struct page *page[1]; |
1806 | unsigned long addr; | |
1807 | int npages; | |
1808 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1809 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1810 | |
777b3f49 MT |
1811 | addr = gfn_to_hva(kvm, gfn); |
1812 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1813 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1814 | |
1815 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1816 | NULL); | |
1817 | if (unlikely(npages != 1)) | |
e4a533a4 | 1818 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1819 | |
1820 | vmf->page = page[0]; | |
e4a533a4 | 1821 | return 0; |
f17abe9a AK |
1822 | } |
1823 | ||
f0f37e2f | 1824 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 1825 | .fault = kvm_vm_fault, |
f17abe9a AK |
1826 | }; |
1827 | ||
1828 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1829 | { | |
1830 | vma->vm_ops = &kvm_vm_vm_ops; | |
1831 | return 0; | |
1832 | } | |
1833 | ||
3d3aab1b | 1834 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1835 | .release = kvm_vm_release, |
1836 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
1837 | #ifdef CONFIG_COMPAT |
1838 | .compat_ioctl = kvm_vm_compat_ioctl, | |
1839 | #endif | |
f17abe9a | 1840 | .mmap = kvm_vm_mmap, |
6038f373 | 1841 | .llseek = noop_llseek, |
f17abe9a AK |
1842 | }; |
1843 | ||
1844 | static int kvm_dev_ioctl_create_vm(void) | |
1845 | { | |
6ce5a090 | 1846 | int fd, r; |
f17abe9a AK |
1847 | struct kvm *kvm; |
1848 | ||
f17abe9a | 1849 | kvm = kvm_create_vm(); |
d6d28168 AK |
1850 | if (IS_ERR(kvm)) |
1851 | return PTR_ERR(kvm); | |
6ce5a090 TY |
1852 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1853 | r = kvm_coalesced_mmio_init(kvm); | |
1854 | if (r < 0) { | |
1855 | kvm_put_kvm(kvm); | |
1856 | return r; | |
1857 | } | |
1858 | #endif | |
628ff7c1 | 1859 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2030a42c | 1860 | if (fd < 0) |
66c0b394 | 1861 | kvm_put_kvm(kvm); |
f17abe9a | 1862 | |
f17abe9a | 1863 | return fd; |
f17abe9a AK |
1864 | } |
1865 | ||
1a811b61 AK |
1866 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
1867 | { | |
1868 | switch (arg) { | |
ca9edaee | 1869 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 1870 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 1871 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
1872 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1873 | case KVM_CAP_SET_BOOT_CPU_ID: | |
1874 | #endif | |
a9c7399d | 1875 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 1876 | return 1; |
399ec807 AK |
1877 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
1878 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 1879 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 1880 | #endif |
1a811b61 AK |
1881 | default: |
1882 | break; | |
1883 | } | |
1884 | return kvm_dev_ioctl_check_extension(arg); | |
1885 | } | |
1886 | ||
f17abe9a AK |
1887 | static long kvm_dev_ioctl(struct file *filp, |
1888 | unsigned int ioctl, unsigned long arg) | |
1889 | { | |
07c45a36 | 1890 | long r = -EINVAL; |
f17abe9a AK |
1891 | |
1892 | switch (ioctl) { | |
1893 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1894 | r = -EINVAL; |
1895 | if (arg) | |
1896 | goto out; | |
f17abe9a AK |
1897 | r = KVM_API_VERSION; |
1898 | break; | |
1899 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1900 | r = -EINVAL; |
1901 | if (arg) | |
1902 | goto out; | |
f17abe9a AK |
1903 | r = kvm_dev_ioctl_create_vm(); |
1904 | break; | |
018d00d2 | 1905 | case KVM_CHECK_EXTENSION: |
1a811b61 | 1906 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 1907 | break; |
07c45a36 AK |
1908 | case KVM_GET_VCPU_MMAP_SIZE: |
1909 | r = -EINVAL; | |
1910 | if (arg) | |
1911 | goto out; | |
adb1ff46 AK |
1912 | r = PAGE_SIZE; /* struct kvm_run */ |
1913 | #ifdef CONFIG_X86 | |
1914 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
1915 | #endif |
1916 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1917 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 1918 | #endif |
07c45a36 | 1919 | break; |
d4c9ff2d FEL |
1920 | case KVM_TRACE_ENABLE: |
1921 | case KVM_TRACE_PAUSE: | |
1922 | case KVM_TRACE_DISABLE: | |
2023a29c | 1923 | r = -EOPNOTSUPP; |
d4c9ff2d | 1924 | break; |
6aa8b732 | 1925 | default: |
043405e1 | 1926 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1927 | } |
1928 | out: | |
1929 | return r; | |
1930 | } | |
1931 | ||
6aa8b732 | 1932 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1933 | .unlocked_ioctl = kvm_dev_ioctl, |
1934 | .compat_ioctl = kvm_dev_ioctl, | |
6038f373 | 1935 | .llseek = noop_llseek, |
6aa8b732 AK |
1936 | }; |
1937 | ||
1938 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1939 | KVM_MINOR, |
6aa8b732 AK |
1940 | "kvm", |
1941 | &kvm_chardev_ops, | |
1942 | }; | |
1943 | ||
1b6c0168 AK |
1944 | static void hardware_enable(void *junk) |
1945 | { | |
1946 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 1947 | int r; |
1b6c0168 | 1948 | |
7f59f492 | 1949 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1950 | return; |
10474ae8 | 1951 | |
7f59f492 | 1952 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
1953 | |
1954 | r = kvm_arch_hardware_enable(NULL); | |
1955 | ||
1956 | if (r) { | |
1957 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
1958 | atomic_inc(&hardware_enable_failed); | |
1959 | printk(KERN_INFO "kvm: enabling virtualization on " | |
1960 | "CPU%d failed\n", cpu); | |
1961 | } | |
1b6c0168 AK |
1962 | } |
1963 | ||
1964 | static void hardware_disable(void *junk) | |
1965 | { | |
1966 | int cpu = raw_smp_processor_id(); | |
1967 | ||
7f59f492 | 1968 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1969 | return; |
7f59f492 | 1970 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 1971 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1972 | } |
1973 | ||
10474ae8 AG |
1974 | static void hardware_disable_all_nolock(void) |
1975 | { | |
1976 | BUG_ON(!kvm_usage_count); | |
1977 | ||
1978 | kvm_usage_count--; | |
1979 | if (!kvm_usage_count) | |
1980 | on_each_cpu(hardware_disable, NULL, 1); | |
1981 | } | |
1982 | ||
1983 | static void hardware_disable_all(void) | |
1984 | { | |
1985 | spin_lock(&kvm_lock); | |
1986 | hardware_disable_all_nolock(); | |
1987 | spin_unlock(&kvm_lock); | |
1988 | } | |
1989 | ||
1990 | static int hardware_enable_all(void) | |
1991 | { | |
1992 | int r = 0; | |
1993 | ||
1994 | spin_lock(&kvm_lock); | |
1995 | ||
1996 | kvm_usage_count++; | |
1997 | if (kvm_usage_count == 1) { | |
1998 | atomic_set(&hardware_enable_failed, 0); | |
1999 | on_each_cpu(hardware_enable, NULL, 1); | |
2000 | ||
2001 | if (atomic_read(&hardware_enable_failed)) { | |
2002 | hardware_disable_all_nolock(); | |
2003 | r = -EBUSY; | |
2004 | } | |
2005 | } | |
2006 | ||
2007 | spin_unlock(&kvm_lock); | |
2008 | ||
2009 | return r; | |
2010 | } | |
2011 | ||
774c47f1 AK |
2012 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2013 | void *v) | |
2014 | { | |
2015 | int cpu = (long)v; | |
2016 | ||
10474ae8 AG |
2017 | if (!kvm_usage_count) |
2018 | return NOTIFY_OK; | |
2019 | ||
1a6f4d7f | 2020 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2021 | switch (val) { |
cec9ad27 | 2022 | case CPU_DYING: |
6ec8a856 AK |
2023 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2024 | cpu); | |
2025 | hardware_disable(NULL); | |
2026 | break; | |
da908f2f | 2027 | case CPU_STARTING: |
43934a38 JK |
2028 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2029 | cpu); | |
ca84d1a2 | 2030 | spin_lock(&kvm_lock); |
da908f2f | 2031 | hardware_enable(NULL); |
ca84d1a2 | 2032 | spin_unlock(&kvm_lock); |
774c47f1 AK |
2033 | break; |
2034 | } | |
2035 | return NOTIFY_OK; | |
2036 | } | |
2037 | ||
4ecac3fd AK |
2038 | |
2039 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
2040 | { | |
ca242ac9 | 2041 | if (kvm_rebooting) { |
4ecac3fd | 2042 | /* spin while reset goes on */ |
ca242ac9 | 2043 | local_irq_enable(); |
4ecac3fd | 2044 | while (true) |
624d84cf | 2045 | cpu_relax(); |
ca242ac9 | 2046 | } |
4ecac3fd AK |
2047 | /* Fault while not rebooting. We want the trace. */ |
2048 | BUG(); | |
2049 | } | |
2050 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
2051 | ||
9a2b85c6 | 2052 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2053 | void *v) |
9a2b85c6 | 2054 | { |
8e1c1815 SY |
2055 | /* |
2056 | * Some (well, at least mine) BIOSes hang on reboot if | |
2057 | * in vmx root mode. | |
2058 | * | |
2059 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2060 | */ | |
2061 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2062 | kvm_rebooting = true; | |
2063 | on_each_cpu(hardware_disable, NULL, 1); | |
9a2b85c6 RR |
2064 | return NOTIFY_OK; |
2065 | } | |
2066 | ||
2067 | static struct notifier_block kvm_reboot_notifier = { | |
2068 | .notifier_call = kvm_reboot, | |
2069 | .priority = 0, | |
2070 | }; | |
2071 | ||
e93f8a0f | 2072 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
2073 | { |
2074 | int i; | |
2075 | ||
2076 | for (i = 0; i < bus->dev_count; i++) { | |
2077 | struct kvm_io_device *pos = bus->devs[i]; | |
2078 | ||
2079 | kvm_iodevice_destructor(pos); | |
2080 | } | |
e93f8a0f | 2081 | kfree(bus); |
2eeb2e94 GH |
2082 | } |
2083 | ||
bda9020e | 2084 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2085 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2086 | int len, const void *val) |
2eeb2e94 GH |
2087 | { |
2088 | int i; | |
90d83dc3 LJ |
2089 | struct kvm_io_bus *bus; |
2090 | ||
2091 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
bda9020e MT |
2092 | for (i = 0; i < bus->dev_count; i++) |
2093 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
2094 | return 0; | |
2095 | return -EOPNOTSUPP; | |
2096 | } | |
2eeb2e94 | 2097 | |
bda9020e | 2098 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2099 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2100 | int len, void *val) | |
bda9020e MT |
2101 | { |
2102 | int i; | |
90d83dc3 | 2103 | struct kvm_io_bus *bus; |
e93f8a0f | 2104 | |
90d83dc3 | 2105 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
bda9020e MT |
2106 | for (i = 0; i < bus->dev_count; i++) |
2107 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
2108 | return 0; | |
2109 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
2110 | } |
2111 | ||
79fac95e | 2112 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2113 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2114 | struct kvm_io_device *dev) | |
6c474694 | 2115 | { |
e93f8a0f | 2116 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2117 | |
e93f8a0f | 2118 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2119 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2120 | return -ENOSPC; | |
2eeb2e94 | 2121 | |
e93f8a0f MT |
2122 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2123 | if (!new_bus) | |
2124 | return -ENOMEM; | |
2125 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2126 | new_bus->devs[new_bus->dev_count++] = dev; | |
2127 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2128 | synchronize_srcu_expedited(&kvm->srcu); | |
2129 | kfree(bus); | |
090b7aff GH |
2130 | |
2131 | return 0; | |
2132 | } | |
2133 | ||
79fac95e | 2134 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2135 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2136 | struct kvm_io_device *dev) | |
090b7aff | 2137 | { |
e93f8a0f MT |
2138 | int i, r; |
2139 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2140 | |
e93f8a0f MT |
2141 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2142 | if (!new_bus) | |
2143 | return -ENOMEM; | |
090b7aff | 2144 | |
e93f8a0f MT |
2145 | bus = kvm->buses[bus_idx]; |
2146 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2147 | ||
2148 | r = -ENOENT; | |
2149 | for (i = 0; i < new_bus->dev_count; i++) | |
2150 | if (new_bus->devs[i] == dev) { | |
2151 | r = 0; | |
2152 | new_bus->devs[i] = new_bus->devs[--new_bus->dev_count]; | |
090b7aff GH |
2153 | break; |
2154 | } | |
e93f8a0f MT |
2155 | |
2156 | if (r) { | |
2157 | kfree(new_bus); | |
2158 | return r; | |
2159 | } | |
2160 | ||
2161 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2162 | synchronize_srcu_expedited(&kvm->srcu); | |
2163 | kfree(bus); | |
2164 | return r; | |
2eeb2e94 GH |
2165 | } |
2166 | ||
774c47f1 AK |
2167 | static struct notifier_block kvm_cpu_notifier = { |
2168 | .notifier_call = kvm_cpu_hotplug, | |
774c47f1 AK |
2169 | }; |
2170 | ||
8b88b099 | 2171 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2172 | { |
2173 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2174 | struct kvm *kvm; |
2175 | ||
8b88b099 | 2176 | *val = 0; |
ba1389b7 AK |
2177 | spin_lock(&kvm_lock); |
2178 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2179 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2180 | spin_unlock(&kvm_lock); |
8b88b099 | 2181 | return 0; |
ba1389b7 AK |
2182 | } |
2183 | ||
2184 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2185 | ||
8b88b099 | 2186 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2187 | { |
2188 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2189 | struct kvm *kvm; |
2190 | struct kvm_vcpu *vcpu; | |
2191 | int i; | |
2192 | ||
8b88b099 | 2193 | *val = 0; |
1165f5fe AK |
2194 | spin_lock(&kvm_lock); |
2195 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
2196 | kvm_for_each_vcpu(i, vcpu, kvm) |
2197 | *val += *(u32 *)((void *)vcpu + offset); | |
2198 | ||
1165f5fe | 2199 | spin_unlock(&kvm_lock); |
8b88b099 | 2200 | return 0; |
1165f5fe AK |
2201 | } |
2202 | ||
ba1389b7 AK |
2203 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2204 | ||
828c0950 | 2205 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2206 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2207 | [KVM_STAT_VM] = &vm_stat_fops, | |
2208 | }; | |
1165f5fe | 2209 | |
a16b043c | 2210 | static void kvm_init_debug(void) |
6aa8b732 AK |
2211 | { |
2212 | struct kvm_stats_debugfs_item *p; | |
2213 | ||
76f7c879 | 2214 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2215 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2216 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2217 | (void *)(long)p->offset, |
ba1389b7 | 2218 | stat_fops[p->kind]); |
6aa8b732 AK |
2219 | } |
2220 | ||
2221 | static void kvm_exit_debug(void) | |
2222 | { | |
2223 | struct kvm_stats_debugfs_item *p; | |
2224 | ||
2225 | for (p = debugfs_entries; p->name; ++p) | |
2226 | debugfs_remove(p->dentry); | |
76f7c879 | 2227 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2228 | } |
2229 | ||
59ae6c6b AK |
2230 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2231 | { | |
10474ae8 AG |
2232 | if (kvm_usage_count) |
2233 | hardware_disable(NULL); | |
59ae6c6b AK |
2234 | return 0; |
2235 | } | |
2236 | ||
2237 | static int kvm_resume(struct sys_device *dev) | |
2238 | { | |
ca84d1a2 ZA |
2239 | if (kvm_usage_count) { |
2240 | WARN_ON(spin_is_locked(&kvm_lock)); | |
10474ae8 | 2241 | hardware_enable(NULL); |
ca84d1a2 | 2242 | } |
59ae6c6b AK |
2243 | return 0; |
2244 | } | |
2245 | ||
2246 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2247 | .name = "kvm", |
59ae6c6b AK |
2248 | .suspend = kvm_suspend, |
2249 | .resume = kvm_resume, | |
2250 | }; | |
2251 | ||
2252 | static struct sys_device kvm_sysdev = { | |
2253 | .id = 0, | |
2254 | .cls = &kvm_sysdev_class, | |
2255 | }; | |
2256 | ||
cea7bb21 | 2257 | struct page *bad_page; |
35149e21 | 2258 | pfn_t bad_pfn; |
6aa8b732 | 2259 | |
15ad7146 AK |
2260 | static inline |
2261 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2262 | { | |
2263 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2264 | } | |
2265 | ||
2266 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2267 | { | |
2268 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2269 | ||
e9b11c17 | 2270 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2271 | } |
2272 | ||
2273 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2274 | struct task_struct *next) | |
2275 | { | |
2276 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2277 | ||
e9b11c17 | 2278 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2279 | } |
2280 | ||
0ee75bea | 2281 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2282 | struct module *module) |
6aa8b732 AK |
2283 | { |
2284 | int r; | |
002c7f7c | 2285 | int cpu; |
6aa8b732 | 2286 | |
f8c16bba ZX |
2287 | r = kvm_arch_init(opaque); |
2288 | if (r) | |
d2308784 | 2289 | goto out_fail; |
cb498ea2 ZX |
2290 | |
2291 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2292 | ||
2293 | if (bad_page == NULL) { | |
2294 | r = -ENOMEM; | |
2295 | goto out; | |
2296 | } | |
2297 | ||
35149e21 AL |
2298 | bad_pfn = page_to_pfn(bad_page); |
2299 | ||
bf998156 HY |
2300 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2301 | ||
2302 | if (hwpoison_page == NULL) { | |
2303 | r = -ENOMEM; | |
2304 | goto out_free_0; | |
2305 | } | |
2306 | ||
2307 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2308 | ||
edba23e5 GN |
2309 | fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2310 | ||
2311 | if (fault_page == NULL) { | |
2312 | r = -ENOMEM; | |
2313 | goto out_free_0; | |
2314 | } | |
2315 | ||
2316 | fault_pfn = page_to_pfn(fault_page); | |
2317 | ||
8437a617 | 2318 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2319 | r = -ENOMEM; |
2320 | goto out_free_0; | |
2321 | } | |
2322 | ||
e9b11c17 | 2323 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2324 | if (r < 0) |
7f59f492 | 2325 | goto out_free_0a; |
6aa8b732 | 2326 | |
002c7f7c YS |
2327 | for_each_online_cpu(cpu) { |
2328 | smp_call_function_single(cpu, | |
e9b11c17 | 2329 | kvm_arch_check_processor_compat, |
8691e5a8 | 2330 | &r, 1); |
002c7f7c | 2331 | if (r < 0) |
d2308784 | 2332 | goto out_free_1; |
002c7f7c YS |
2333 | } |
2334 | ||
774c47f1 AK |
2335 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2336 | if (r) | |
d2308784 | 2337 | goto out_free_2; |
6aa8b732 AK |
2338 | register_reboot_notifier(&kvm_reboot_notifier); |
2339 | ||
59ae6c6b AK |
2340 | r = sysdev_class_register(&kvm_sysdev_class); |
2341 | if (r) | |
d2308784 | 2342 | goto out_free_3; |
59ae6c6b AK |
2343 | |
2344 | r = sysdev_register(&kvm_sysdev); | |
2345 | if (r) | |
d2308784 | 2346 | goto out_free_4; |
59ae6c6b | 2347 | |
c16f862d | 2348 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2349 | if (!vcpu_align) |
2350 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2351 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2352 | 0, NULL); |
c16f862d RR |
2353 | if (!kvm_vcpu_cache) { |
2354 | r = -ENOMEM; | |
d2308784 | 2355 | goto out_free_5; |
c16f862d RR |
2356 | } |
2357 | ||
af585b92 GN |
2358 | r = kvm_async_pf_init(); |
2359 | if (r) | |
2360 | goto out_free; | |
2361 | ||
6aa8b732 | 2362 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2363 | kvm_vm_fops.owner = module; |
2364 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2365 | |
2366 | r = misc_register(&kvm_dev); | |
2367 | if (r) { | |
d77c26fc | 2368 | printk(KERN_ERR "kvm: misc device register failed\n"); |
af585b92 | 2369 | goto out_unreg; |
6aa8b732 AK |
2370 | } |
2371 | ||
15ad7146 AK |
2372 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2373 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2374 | ||
0ea4ed8e DW |
2375 | kvm_init_debug(); |
2376 | ||
c7addb90 | 2377 | return 0; |
6aa8b732 | 2378 | |
af585b92 GN |
2379 | out_unreg: |
2380 | kvm_async_pf_deinit(); | |
6aa8b732 | 2381 | out_free: |
c16f862d | 2382 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2383 | out_free_5: |
59ae6c6b | 2384 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2385 | out_free_4: |
59ae6c6b | 2386 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2387 | out_free_3: |
6aa8b732 | 2388 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2389 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2390 | out_free_2: |
d2308784 | 2391 | out_free_1: |
e9b11c17 | 2392 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2393 | out_free_0a: |
2394 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2395 | out_free_0: |
edba23e5 GN |
2396 | if (fault_page) |
2397 | __free_page(fault_page); | |
bf998156 HY |
2398 | if (hwpoison_page) |
2399 | __free_page(hwpoison_page); | |
d2308784 | 2400 | __free_page(bad_page); |
ca45aaae | 2401 | out: |
f8c16bba | 2402 | kvm_arch_exit(); |
d2308784 | 2403 | out_fail: |
6aa8b732 AK |
2404 | return r; |
2405 | } | |
cb498ea2 | 2406 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2407 | |
cb498ea2 | 2408 | void kvm_exit(void) |
6aa8b732 | 2409 | { |
0ea4ed8e | 2410 | kvm_exit_debug(); |
6aa8b732 | 2411 | misc_deregister(&kvm_dev); |
c16f862d | 2412 | kmem_cache_destroy(kvm_vcpu_cache); |
af585b92 | 2413 | kvm_async_pf_deinit(); |
59ae6c6b AK |
2414 | sysdev_unregister(&kvm_sysdev); |
2415 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2416 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2417 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2418 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2419 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2420 | kvm_arch_exit(); |
7f59f492 | 2421 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2422 | __free_page(hwpoison_page); |
cea7bb21 | 2423 | __free_page(bad_page); |
6aa8b732 | 2424 | } |
cb498ea2 | 2425 | EXPORT_SYMBOL_GPL(kvm_exit); |