]>
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> | |
fb3600cc | 33 | #include <linux/syscore_ops.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> |
743eeb0b SL |
50 | #include <linux/sort.h> |
51 | #include <linux/bsearch.h> | |
6aa8b732 | 52 | |
e495606d | 53 | #include <asm/processor.h> |
e495606d AK |
54 | #include <asm/io.h> |
55 | #include <asm/uaccess.h> | |
3e021bf5 | 56 | #include <asm/pgtable.h> |
6aa8b732 | 57 | |
5f94c174 | 58 | #include "coalesced_mmio.h" |
af585b92 | 59 | #include "async_pf.h" |
5f94c174 | 60 | |
229456fc MT |
61 | #define CREATE_TRACE_POINTS |
62 | #include <trace/events/kvm.h> | |
63 | ||
6aa8b732 AK |
64 | MODULE_AUTHOR("Qumranet"); |
65 | MODULE_LICENSE("GPL"); | |
66 | ||
fa40a821 MT |
67 | /* |
68 | * Ordering of locks: | |
69 | * | |
fae3a353 | 70 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
71 | */ |
72 | ||
e935b837 | 73 | DEFINE_RAW_SPINLOCK(kvm_lock); |
e9b11c17 | 74 | LIST_HEAD(vm_list); |
133de902 | 75 | |
7f59f492 | 76 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
77 | static int kvm_usage_count = 0; |
78 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 79 | |
c16f862d RR |
80 | struct kmem_cache *kvm_vcpu_cache; |
81 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 82 | |
15ad7146 AK |
83 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
84 | ||
76f7c879 | 85 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 86 | |
bccf2150 AK |
87 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
88 | unsigned long arg); | |
1dda606c AG |
89 | #ifdef CONFIG_COMPAT |
90 | static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl, | |
91 | unsigned long arg); | |
92 | #endif | |
10474ae8 AG |
93 | static int hardware_enable_all(void); |
94 | static void hardware_disable_all(void); | |
bccf2150 | 95 | |
e93f8a0f MT |
96 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
97 | ||
b7c4145b AK |
98 | bool kvm_rebooting; |
99 | EXPORT_SYMBOL_GPL(kvm_rebooting); | |
4ecac3fd | 100 | |
54dee993 MT |
101 | static bool largepages_enabled = true; |
102 | ||
a2766325 | 103 | bool kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 104 | { |
a2766325 XG |
105 | if (is_error_pfn(pfn)) |
106 | return false; | |
107 | ||
fc5659c8 | 108 | if (pfn_valid(pfn)) { |
22e5c47e | 109 | int reserved; |
936a5fe6 | 110 | struct page *tail = pfn_to_page(pfn); |
22e5c47e AA |
111 | struct page *head = compound_trans_head(tail); |
112 | reserved = PageReserved(head); | |
936a5fe6 | 113 | if (head != tail) { |
936a5fe6 | 114 | /* |
22e5c47e AA |
115 | * "head" is not a dangling pointer |
116 | * (compound_trans_head takes care of that) | |
117 | * but the hugepage may have been splitted | |
118 | * from under us (and we may not hold a | |
119 | * reference count on the head page so it can | |
120 | * be reused before we run PageReferenced), so | |
121 | * we've to check PageTail before returning | |
122 | * what we just read. | |
936a5fe6 | 123 | */ |
22e5c47e AA |
124 | smp_rmb(); |
125 | if (PageTail(tail)) | |
126 | return reserved; | |
936a5fe6 AA |
127 | } |
128 | return PageReserved(tail); | |
fc5659c8 | 129 | } |
cbff90a7 BAY |
130 | |
131 | return true; | |
132 | } | |
133 | ||
bccf2150 AK |
134 | /* |
135 | * Switches to specified vcpu, until a matching vcpu_put() | |
136 | */ | |
313a3dc7 | 137 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 138 | { |
15ad7146 AK |
139 | int cpu; |
140 | ||
bccf2150 | 141 | mutex_lock(&vcpu->mutex); |
34bb10b7 RR |
142 | if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) { |
143 | /* The thread running this VCPU changed. */ | |
144 | struct pid *oldpid = vcpu->pid; | |
145 | struct pid *newpid = get_task_pid(current, PIDTYPE_PID); | |
146 | rcu_assign_pointer(vcpu->pid, newpid); | |
147 | synchronize_rcu(); | |
148 | put_pid(oldpid); | |
149 | } | |
15ad7146 AK |
150 | cpu = get_cpu(); |
151 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 152 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 153 | put_cpu(); |
6aa8b732 AK |
154 | } |
155 | ||
313a3dc7 | 156 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 157 | { |
15ad7146 | 158 | preempt_disable(); |
313a3dc7 | 159 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
160 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
161 | preempt_enable(); | |
6aa8b732 AK |
162 | mutex_unlock(&vcpu->mutex); |
163 | } | |
164 | ||
d9e368d6 AK |
165 | static void ack_flush(void *_completed) |
166 | { | |
d9e368d6 AK |
167 | } |
168 | ||
49846896 | 169 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 170 | { |
597a5f55 | 171 | int i, cpu, me; |
6ef7a1bc RR |
172 | cpumask_var_t cpus; |
173 | bool called = true; | |
d9e368d6 | 174 | struct kvm_vcpu *vcpu; |
d9e368d6 | 175 | |
79f55997 | 176 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 177 | |
3cba4130 | 178 | me = get_cpu(); |
988a2cae | 179 | kvm_for_each_vcpu(i, vcpu, kvm) { |
3cba4130 | 180 | kvm_make_request(req, vcpu); |
d9e368d6 | 181 | cpu = vcpu->cpu; |
6b7e2d09 XG |
182 | |
183 | /* Set ->requests bit before we read ->mode */ | |
184 | smp_mb(); | |
185 | ||
186 | if (cpus != NULL && cpu != -1 && cpu != me && | |
187 | kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE) | |
6ef7a1bc | 188 | cpumask_set_cpu(cpu, cpus); |
49846896 | 189 | } |
6ef7a1bc RR |
190 | if (unlikely(cpus == NULL)) |
191 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
192 | else if (!cpumask_empty(cpus)) | |
193 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
194 | else | |
195 | called = false; | |
3cba4130 | 196 | put_cpu(); |
6ef7a1bc | 197 | free_cpumask_var(cpus); |
49846896 | 198 | return called; |
d9e368d6 AK |
199 | } |
200 | ||
49846896 | 201 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 202 | { |
bec87d6e | 203 | long dirty_count = kvm->tlbs_dirty; |
a4ee1ca4 XG |
204 | |
205 | smp_mb(); | |
49846896 RR |
206 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
207 | ++kvm->stat.remote_tlb_flush; | |
a4ee1ca4 | 208 | cmpxchg(&kvm->tlbs_dirty, dirty_count, 0); |
2e53d63a MT |
209 | } |
210 | ||
49846896 RR |
211 | void kvm_reload_remote_mmus(struct kvm *kvm) |
212 | { | |
213 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
214 | } | |
2e53d63a | 215 | |
fb3f0f51 RR |
216 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
217 | { | |
218 | struct page *page; | |
219 | int r; | |
220 | ||
221 | mutex_init(&vcpu->mutex); | |
222 | vcpu->cpu = -1; | |
fb3f0f51 RR |
223 | vcpu->kvm = kvm; |
224 | vcpu->vcpu_id = id; | |
34bb10b7 | 225 | vcpu->pid = NULL; |
b6958ce4 | 226 | init_waitqueue_head(&vcpu->wq); |
af585b92 | 227 | kvm_async_pf_vcpu_init(vcpu); |
fb3f0f51 RR |
228 | |
229 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
230 | if (!page) { | |
231 | r = -ENOMEM; | |
232 | goto fail; | |
233 | } | |
234 | vcpu->run = page_address(page); | |
235 | ||
4c088493 R |
236 | kvm_vcpu_set_in_spin_loop(vcpu, false); |
237 | kvm_vcpu_set_dy_eligible(vcpu, false); | |
238 | ||
e9b11c17 | 239 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 240 | if (r < 0) |
e9b11c17 | 241 | goto fail_free_run; |
fb3f0f51 RR |
242 | return 0; |
243 | ||
fb3f0f51 RR |
244 | fail_free_run: |
245 | free_page((unsigned long)vcpu->run); | |
246 | fail: | |
76fafa5e | 247 | return r; |
fb3f0f51 RR |
248 | } |
249 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
250 | ||
251 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
252 | { | |
34bb10b7 | 253 | put_pid(vcpu->pid); |
e9b11c17 | 254 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
255 | free_page((unsigned long)vcpu->run); |
256 | } | |
257 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
258 | ||
e930bffe AA |
259 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
260 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
261 | { | |
262 | return container_of(mn, struct kvm, mmu_notifier); | |
263 | } | |
264 | ||
265 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
266 | struct mm_struct *mm, | |
267 | unsigned long address) | |
268 | { | |
269 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 270 | int need_tlb_flush, idx; |
e930bffe AA |
271 | |
272 | /* | |
273 | * When ->invalidate_page runs, the linux pte has been zapped | |
274 | * already but the page is still allocated until | |
275 | * ->invalidate_page returns. So if we increase the sequence | |
276 | * here the kvm page fault will notice if the spte can't be | |
277 | * established because the page is going to be freed. If | |
278 | * instead the kvm page fault establishes the spte before | |
279 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
280 | * before returning. | |
281 | * | |
282 | * The sequence increase only need to be seen at spin_unlock | |
283 | * time, and not at spin_lock time. | |
284 | * | |
285 | * Increasing the sequence after the spin_unlock would be | |
286 | * unsafe because the kvm page fault could then establish the | |
287 | * pte after kvm_unmap_hva returned, without noticing the page | |
288 | * is going to be freed. | |
289 | */ | |
bc6678a3 | 290 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe | 291 | spin_lock(&kvm->mmu_lock); |
565f3be2 | 292 | |
e930bffe | 293 | kvm->mmu_notifier_seq++; |
a4ee1ca4 | 294 | need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty; |
e930bffe AA |
295 | /* we've to flush the tlb before the pages can be freed */ |
296 | if (need_tlb_flush) | |
297 | kvm_flush_remote_tlbs(kvm); | |
298 | ||
565f3be2 TY |
299 | spin_unlock(&kvm->mmu_lock); |
300 | srcu_read_unlock(&kvm->srcu, idx); | |
e930bffe AA |
301 | } |
302 | ||
3da0dd43 IE |
303 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
304 | struct mm_struct *mm, | |
305 | unsigned long address, | |
306 | pte_t pte) | |
307 | { | |
308 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 309 | int idx; |
3da0dd43 | 310 | |
bc6678a3 | 311 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
312 | spin_lock(&kvm->mmu_lock); |
313 | kvm->mmu_notifier_seq++; | |
314 | kvm_set_spte_hva(kvm, address, pte); | |
315 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 316 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
317 | } |
318 | ||
e930bffe AA |
319 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
320 | struct mm_struct *mm, | |
321 | unsigned long start, | |
322 | unsigned long end) | |
323 | { | |
324 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 325 | int need_tlb_flush = 0, idx; |
e930bffe | 326 | |
bc6678a3 | 327 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
328 | spin_lock(&kvm->mmu_lock); |
329 | /* | |
330 | * The count increase must become visible at unlock time as no | |
331 | * spte can be established without taking the mmu_lock and | |
332 | * count is also read inside the mmu_lock critical section. | |
333 | */ | |
334 | kvm->mmu_notifier_count++; | |
b3ae2096 | 335 | need_tlb_flush = kvm_unmap_hva_range(kvm, start, end); |
a4ee1ca4 | 336 | need_tlb_flush |= kvm->tlbs_dirty; |
e930bffe AA |
337 | /* we've to flush the tlb before the pages can be freed */ |
338 | if (need_tlb_flush) | |
339 | kvm_flush_remote_tlbs(kvm); | |
565f3be2 TY |
340 | |
341 | spin_unlock(&kvm->mmu_lock); | |
342 | srcu_read_unlock(&kvm->srcu, idx); | |
e930bffe AA |
343 | } |
344 | ||
345 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
346 | struct mm_struct *mm, | |
347 | unsigned long start, | |
348 | unsigned long end) | |
349 | { | |
350 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
351 | ||
352 | spin_lock(&kvm->mmu_lock); | |
353 | /* | |
354 | * This sequence increase will notify the kvm page fault that | |
355 | * the page that is going to be mapped in the spte could have | |
356 | * been freed. | |
357 | */ | |
358 | kvm->mmu_notifier_seq++; | |
a355aa54 | 359 | smp_wmb(); |
e930bffe AA |
360 | /* |
361 | * The above sequence increase must be visible before the | |
a355aa54 PM |
362 | * below count decrease, which is ensured by the smp_wmb above |
363 | * in conjunction with the smp_rmb in mmu_notifier_retry(). | |
e930bffe AA |
364 | */ |
365 | kvm->mmu_notifier_count--; | |
366 | spin_unlock(&kvm->mmu_lock); | |
367 | ||
368 | BUG_ON(kvm->mmu_notifier_count < 0); | |
369 | } | |
370 | ||
371 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
372 | struct mm_struct *mm, | |
373 | unsigned long address) | |
374 | { | |
375 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 376 | int young, idx; |
e930bffe | 377 | |
bc6678a3 | 378 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe | 379 | spin_lock(&kvm->mmu_lock); |
e930bffe | 380 | |
565f3be2 | 381 | young = kvm_age_hva(kvm, address); |
e930bffe AA |
382 | if (young) |
383 | kvm_flush_remote_tlbs(kvm); | |
384 | ||
565f3be2 TY |
385 | spin_unlock(&kvm->mmu_lock); |
386 | srcu_read_unlock(&kvm->srcu, idx); | |
387 | ||
e930bffe AA |
388 | return young; |
389 | } | |
390 | ||
8ee53820 AA |
391 | static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn, |
392 | struct mm_struct *mm, | |
393 | unsigned long address) | |
394 | { | |
395 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
396 | int young, idx; | |
397 | ||
398 | idx = srcu_read_lock(&kvm->srcu); | |
399 | spin_lock(&kvm->mmu_lock); | |
400 | young = kvm_test_age_hva(kvm, address); | |
401 | spin_unlock(&kvm->mmu_lock); | |
402 | srcu_read_unlock(&kvm->srcu, idx); | |
403 | ||
404 | return young; | |
405 | } | |
406 | ||
85db06e5 MT |
407 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
408 | struct mm_struct *mm) | |
409 | { | |
410 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
411 | int idx; |
412 | ||
413 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 414 | kvm_arch_flush_shadow(kvm); |
eda2beda | 415 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
416 | } |
417 | ||
e930bffe AA |
418 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
419 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
420 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
421 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
422 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
8ee53820 | 423 | .test_young = kvm_mmu_notifier_test_young, |
3da0dd43 | 424 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 425 | .release = kvm_mmu_notifier_release, |
e930bffe | 426 | }; |
4c07b0a4 AK |
427 | |
428 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
429 | { | |
430 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
431 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
432 | } | |
433 | ||
434 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
435 | ||
436 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
437 | { | |
438 | return 0; | |
439 | } | |
440 | ||
e930bffe AA |
441 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
442 | ||
bf3e05bc XG |
443 | static void kvm_init_memslots_id(struct kvm *kvm) |
444 | { | |
445 | int i; | |
446 | struct kvm_memslots *slots = kvm->memslots; | |
447 | ||
448 | for (i = 0; i < KVM_MEM_SLOTS_NUM; i++) | |
f85e2cb5 | 449 | slots->id_to_index[i] = slots->memslots[i].id = i; |
bf3e05bc XG |
450 | } |
451 | ||
e08b9637 | 452 | static struct kvm *kvm_create_vm(unsigned long type) |
6aa8b732 | 453 | { |
d89f5eff JK |
454 | int r, i; |
455 | struct kvm *kvm = kvm_arch_alloc_vm(); | |
6aa8b732 | 456 | |
d89f5eff JK |
457 | if (!kvm) |
458 | return ERR_PTR(-ENOMEM); | |
459 | ||
e08b9637 | 460 | r = kvm_arch_init_vm(kvm, type); |
d89f5eff JK |
461 | if (r) |
462 | goto out_err_nodisable; | |
10474ae8 AG |
463 | |
464 | r = hardware_enable_all(); | |
465 | if (r) | |
466 | goto out_err_nodisable; | |
467 | ||
75858a84 AK |
468 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
469 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 470 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 471 | #endif |
6aa8b732 | 472 | |
46a26bf5 MT |
473 | r = -ENOMEM; |
474 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
475 | if (!kvm->memslots) | |
57e7fbee | 476 | goto out_err_nosrcu; |
bf3e05bc | 477 | kvm_init_memslots_id(kvm); |
bc6678a3 | 478 | if (init_srcu_struct(&kvm->srcu)) |
57e7fbee | 479 | goto out_err_nosrcu; |
e93f8a0f MT |
480 | for (i = 0; i < KVM_NR_BUSES; i++) { |
481 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
482 | GFP_KERNEL); | |
57e7fbee | 483 | if (!kvm->buses[i]) |
e93f8a0f | 484 | goto out_err; |
e93f8a0f | 485 | } |
e930bffe | 486 | |
74b5c5bf | 487 | spin_lock_init(&kvm->mmu_lock); |
6d4e4c4f AK |
488 | kvm->mm = current->mm; |
489 | atomic_inc(&kvm->mm->mm_count); | |
d34e6b17 | 490 | kvm_eventfd_init(kvm); |
11ec2804 | 491 | mutex_init(&kvm->lock); |
60eead79 | 492 | mutex_init(&kvm->irq_lock); |
79fac95e | 493 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 494 | atomic_set(&kvm->users_count, 1); |
74b5c5bf MW |
495 | |
496 | r = kvm_init_mmu_notifier(kvm); | |
497 | if (r) | |
498 | goto out_err; | |
499 | ||
e935b837 | 500 | raw_spin_lock(&kvm_lock); |
5e58cfe4 | 501 | list_add(&kvm->vm_list, &vm_list); |
e935b837 | 502 | raw_spin_unlock(&kvm_lock); |
d89f5eff | 503 | |
f17abe9a | 504 | return kvm; |
10474ae8 AG |
505 | |
506 | out_err: | |
57e7fbee JK |
507 | cleanup_srcu_struct(&kvm->srcu); |
508 | out_err_nosrcu: | |
10474ae8 AG |
509 | hardware_disable_all(); |
510 | out_err_nodisable: | |
e93f8a0f MT |
511 | for (i = 0; i < KVM_NR_BUSES; i++) |
512 | kfree(kvm->buses[i]); | |
46a26bf5 | 513 | kfree(kvm->memslots); |
d89f5eff | 514 | kvm_arch_free_vm(kvm); |
10474ae8 | 515 | return ERR_PTR(r); |
f17abe9a AK |
516 | } |
517 | ||
92eca8fa TY |
518 | /* |
519 | * Avoid using vmalloc for a small buffer. | |
520 | * Should not be used when the size is statically known. | |
521 | */ | |
c1a7b32a | 522 | void *kvm_kvzalloc(unsigned long size) |
92eca8fa TY |
523 | { |
524 | if (size > PAGE_SIZE) | |
525 | return vzalloc(size); | |
526 | else | |
527 | return kzalloc(size, GFP_KERNEL); | |
528 | } | |
529 | ||
c1a7b32a | 530 | void kvm_kvfree(const void *addr) |
92eca8fa TY |
531 | { |
532 | if (is_vmalloc_addr(addr)) | |
533 | vfree(addr); | |
534 | else | |
535 | kfree(addr); | |
536 | } | |
537 | ||
a36a57b1 TY |
538 | static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot) |
539 | { | |
540 | if (!memslot->dirty_bitmap) | |
541 | return; | |
542 | ||
92eca8fa | 543 | kvm_kvfree(memslot->dirty_bitmap); |
a36a57b1 TY |
544 | memslot->dirty_bitmap = NULL; |
545 | } | |
546 | ||
6aa8b732 AK |
547 | /* |
548 | * Free any memory in @free but not in @dont. | |
549 | */ | |
550 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
551 | struct kvm_memory_slot *dont) | |
552 | { | |
290fc38d IE |
553 | if (!dont || free->rmap != dont->rmap) |
554 | vfree(free->rmap); | |
6aa8b732 AK |
555 | |
556 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
a36a57b1 | 557 | kvm_destroy_dirty_bitmap(free); |
6aa8b732 | 558 | |
db3fe4eb | 559 | kvm_arch_free_memslot(free, dont); |
05da4558 | 560 | |
6aa8b732 | 561 | free->npages = 0; |
8d4e1288 | 562 | free->rmap = NULL; |
6aa8b732 AK |
563 | } |
564 | ||
d19a9cd2 | 565 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 | 566 | { |
46a26bf5 | 567 | struct kvm_memslots *slots = kvm->memslots; |
be6ba0f0 | 568 | struct kvm_memory_slot *memslot; |
46a26bf5 | 569 | |
be6ba0f0 XG |
570 | kvm_for_each_memslot(memslot, slots) |
571 | kvm_free_physmem_slot(memslot, NULL); | |
6aa8b732 | 572 | |
46a26bf5 | 573 | kfree(kvm->memslots); |
6aa8b732 AK |
574 | } |
575 | ||
f17abe9a AK |
576 | static void kvm_destroy_vm(struct kvm *kvm) |
577 | { | |
e93f8a0f | 578 | int i; |
6d4e4c4f AK |
579 | struct mm_struct *mm = kvm->mm; |
580 | ||
ad8ba2cd | 581 | kvm_arch_sync_events(kvm); |
e935b837 | 582 | raw_spin_lock(&kvm_lock); |
133de902 | 583 | list_del(&kvm->vm_list); |
e935b837 | 584 | raw_spin_unlock(&kvm_lock); |
399ec807 | 585 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
586 | for (i = 0; i < KVM_NR_BUSES; i++) |
587 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 588 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
589 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
590 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
591 | #else |
592 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 593 | #endif |
d19a9cd2 | 594 | kvm_arch_destroy_vm(kvm); |
d89f5eff JK |
595 | kvm_free_physmem(kvm); |
596 | cleanup_srcu_struct(&kvm->srcu); | |
597 | kvm_arch_free_vm(kvm); | |
10474ae8 | 598 | hardware_disable_all(); |
6d4e4c4f | 599 | mmdrop(mm); |
f17abe9a AK |
600 | } |
601 | ||
d39f13b0 IE |
602 | void kvm_get_kvm(struct kvm *kvm) |
603 | { | |
604 | atomic_inc(&kvm->users_count); | |
605 | } | |
606 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
607 | ||
608 | void kvm_put_kvm(struct kvm *kvm) | |
609 | { | |
610 | if (atomic_dec_and_test(&kvm->users_count)) | |
611 | kvm_destroy_vm(kvm); | |
612 | } | |
613 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
614 | ||
615 | ||
f17abe9a AK |
616 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
617 | { | |
618 | struct kvm *kvm = filp->private_data; | |
619 | ||
721eecbf GH |
620 | kvm_irqfd_release(kvm); |
621 | ||
d39f13b0 | 622 | kvm_put_kvm(kvm); |
6aa8b732 AK |
623 | return 0; |
624 | } | |
625 | ||
515a0127 TY |
626 | /* |
627 | * Allocation size is twice as large as the actual dirty bitmap size. | |
93474b25 | 628 | * See x86's kvm_vm_ioctl_get_dirty_log() why this is needed. |
515a0127 | 629 | */ |
a36a57b1 TY |
630 | static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot) |
631 | { | |
189a2f7b | 632 | #ifndef CONFIG_S390 |
515a0127 | 633 | unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot); |
a36a57b1 | 634 | |
92eca8fa | 635 | memslot->dirty_bitmap = kvm_kvzalloc(dirty_bytes); |
a36a57b1 TY |
636 | if (!memslot->dirty_bitmap) |
637 | return -ENOMEM; | |
638 | ||
189a2f7b | 639 | #endif /* !CONFIG_S390 */ |
a36a57b1 TY |
640 | return 0; |
641 | } | |
642 | ||
bf3e05bc XG |
643 | static int cmp_memslot(const void *slot1, const void *slot2) |
644 | { | |
645 | struct kvm_memory_slot *s1, *s2; | |
646 | ||
647 | s1 = (struct kvm_memory_slot *)slot1; | |
648 | s2 = (struct kvm_memory_slot *)slot2; | |
649 | ||
650 | if (s1->npages < s2->npages) | |
651 | return 1; | |
652 | if (s1->npages > s2->npages) | |
653 | return -1; | |
654 | ||
655 | return 0; | |
656 | } | |
657 | ||
658 | /* | |
659 | * Sort the memslots base on its size, so the larger slots | |
660 | * will get better fit. | |
661 | */ | |
662 | static void sort_memslots(struct kvm_memslots *slots) | |
663 | { | |
f85e2cb5 XG |
664 | int i; |
665 | ||
bf3e05bc XG |
666 | sort(slots->memslots, KVM_MEM_SLOTS_NUM, |
667 | sizeof(struct kvm_memory_slot), cmp_memslot, NULL); | |
f85e2cb5 XG |
668 | |
669 | for (i = 0; i < KVM_MEM_SLOTS_NUM; i++) | |
670 | slots->id_to_index[slots->memslots[i].id] = i; | |
bf3e05bc XG |
671 | } |
672 | ||
be593d62 XG |
673 | void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new) |
674 | { | |
675 | if (new) { | |
676 | int id = new->id; | |
28a37544 | 677 | struct kvm_memory_slot *old = id_to_memslot(slots, id); |
bf3e05bc | 678 | unsigned long npages = old->npages; |
be593d62 | 679 | |
28a37544 | 680 | *old = *new; |
bf3e05bc XG |
681 | if (new->npages != npages) |
682 | sort_memslots(slots); | |
be593d62 XG |
683 | } |
684 | ||
685 | slots->generation++; | |
686 | } | |
687 | ||
6aa8b732 AK |
688 | /* |
689 | * Allocate some memory and give it an address in the guest physical address | |
690 | * space. | |
691 | * | |
692 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 693 | * |
10589a46 | 694 | * Must be called holding mmap_sem for write. |
6aa8b732 | 695 | */ |
f78e0e2e SY |
696 | int __kvm_set_memory_region(struct kvm *kvm, |
697 | struct kvm_userspace_memory_region *mem, | |
698 | int user_alloc) | |
6aa8b732 | 699 | { |
8234b22e | 700 | int r; |
6aa8b732 | 701 | gfn_t base_gfn; |
28bcb112 HC |
702 | unsigned long npages; |
703 | unsigned long i; | |
6aa8b732 AK |
704 | struct kvm_memory_slot *memslot; |
705 | struct kvm_memory_slot old, new; | |
bc6678a3 | 706 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
707 | |
708 | r = -EINVAL; | |
709 | /* General sanity checks */ | |
710 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
711 | goto out; | |
712 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
713 | goto out; | |
fa3d315a TY |
714 | /* We can read the guest memory with __xxx_user() later on. */ |
715 | if (user_alloc && | |
716 | ((mem->userspace_addr & (PAGE_SIZE - 1)) || | |
9e3bb6b6 HC |
717 | !access_ok(VERIFY_WRITE, |
718 | (void __user *)(unsigned long)mem->userspace_addr, | |
719 | mem->memory_size))) | |
78749809 | 720 | goto out; |
93a5cef0 | 721 | if (mem->slot >= KVM_MEM_SLOTS_NUM) |
6aa8b732 AK |
722 | goto out; |
723 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
724 | goto out; | |
725 | ||
28a37544 | 726 | memslot = id_to_memslot(kvm->memslots, mem->slot); |
6aa8b732 AK |
727 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
728 | npages = mem->memory_size >> PAGE_SHIFT; | |
729 | ||
660c22c4 TY |
730 | r = -EINVAL; |
731 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
732 | goto out; | |
733 | ||
6aa8b732 AK |
734 | if (!npages) |
735 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
736 | ||
6aa8b732 AK |
737 | new = old = *memslot; |
738 | ||
e36d96f7 | 739 | new.id = mem->slot; |
6aa8b732 AK |
740 | new.base_gfn = base_gfn; |
741 | new.npages = npages; | |
742 | new.flags = mem->flags; | |
743 | ||
744 | /* Disallow changing a memory slot's size. */ | |
745 | r = -EINVAL; | |
746 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 747 | goto out_free; |
6aa8b732 AK |
748 | |
749 | /* Check for overlaps */ | |
750 | r = -EEXIST; | |
751 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 752 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 753 | |
4cd481f6 | 754 | if (s == memslot || !s->npages) |
6aa8b732 AK |
755 | continue; |
756 | if (!((base_gfn + npages <= s->base_gfn) || | |
757 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 758 | goto out_free; |
6aa8b732 | 759 | } |
6aa8b732 | 760 | |
6aa8b732 AK |
761 | /* Free page dirty bitmap if unneeded */ |
762 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 763 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
764 | |
765 | r = -ENOMEM; | |
766 | ||
767 | /* Allocate if a slot is being created */ | |
189a2f7b TY |
768 | if (npages && !old.npages) { |
769 | new.user_alloc = user_alloc; | |
770 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 771 | #ifndef CONFIG_S390 |
26535037 | 772 | new.rmap = vzalloc(npages * sizeof(*new.rmap)); |
290fc38d | 773 | if (!new.rmap) |
f78e0e2e | 774 | goto out_free; |
189a2f7b | 775 | #endif /* not defined CONFIG_S390 */ |
db3fe4eb TY |
776 | if (kvm_arch_create_memslot(&new, npages)) |
777 | goto out_free; | |
6aa8b732 | 778 | } |
ec04b260 | 779 | |
6aa8b732 AK |
780 | /* Allocate page dirty bitmap if needed */ |
781 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
a36a57b1 | 782 | if (kvm_create_dirty_bitmap(&new) < 0) |
f78e0e2e | 783 | goto out_free; |
bc6678a3 | 784 | /* destroy any largepage mappings for dirty tracking */ |
6aa8b732 AK |
785 | } |
786 | ||
bc6678a3 | 787 | if (!npages) { |
28a37544 XG |
788 | struct kvm_memory_slot *slot; |
789 | ||
bc6678a3 | 790 | r = -ENOMEM; |
6da64fdb TM |
791 | slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots), |
792 | GFP_KERNEL); | |
bc6678a3 MT |
793 | if (!slots) |
794 | goto out_free; | |
28a37544 XG |
795 | slot = id_to_memslot(slots, mem->slot); |
796 | slot->flags |= KVM_MEMSLOT_INVALID; | |
797 | ||
be593d62 | 798 | update_memslots(slots, NULL); |
bc6678a3 MT |
799 | |
800 | old_memslots = kvm->memslots; | |
801 | rcu_assign_pointer(kvm->memslots, slots); | |
802 | synchronize_srcu_expedited(&kvm->srcu); | |
803 | /* From this point no new shadow pages pointing to a deleted | |
804 | * memslot will be created. | |
805 | * | |
806 | * validation of sp->gfn happens in: | |
807 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
808 | * - kvm_is_visible_gfn (mmu_check_roots) | |
809 | */ | |
34d4cb8f | 810 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
811 | kfree(old_memslots); |
812 | } | |
34d4cb8f | 813 | |
f7784b8e MT |
814 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
815 | if (r) | |
816 | goto out_free; | |
817 | ||
32f6daad | 818 | /* map/unmap the pages in iommu page table */ |
bc6678a3 MT |
819 | if (npages) { |
820 | r = kvm_iommu_map_pages(kvm, &new); | |
821 | if (r) | |
822 | goto out_free; | |
32f6daad AW |
823 | } else |
824 | kvm_iommu_unmap_pages(kvm, &old); | |
604b38ac | 825 | |
bc6678a3 | 826 | r = -ENOMEM; |
6da64fdb TM |
827 | slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots), |
828 | GFP_KERNEL); | |
bc6678a3 MT |
829 | if (!slots) |
830 | goto out_free; | |
bc6678a3 MT |
831 | |
832 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
833 | if (!npages) { | |
834 | new.rmap = NULL; | |
835 | new.dirty_bitmap = NULL; | |
db3fe4eb | 836 | memset(&new.arch, 0, sizeof(new.arch)); |
bc6678a3 MT |
837 | } |
838 | ||
be593d62 | 839 | update_memslots(slots, &new); |
bc6678a3 MT |
840 | old_memslots = kvm->memslots; |
841 | rcu_assign_pointer(kvm->memslots, slots); | |
842 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 843 | |
f7784b8e | 844 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 845 | |
ce88decf XG |
846 | /* |
847 | * If the new memory slot is created, we need to clear all | |
848 | * mmio sptes. | |
849 | */ | |
850 | if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT) | |
851 | kvm_arch_flush_shadow(kvm); | |
852 | ||
bc6678a3 MT |
853 | kvm_free_physmem_slot(&old, &new); |
854 | kfree(old_memslots); | |
855 | ||
6aa8b732 AK |
856 | return 0; |
857 | ||
f78e0e2e | 858 | out_free: |
6aa8b732 AK |
859 | kvm_free_physmem_slot(&new, &old); |
860 | out: | |
861 | return r; | |
210c7c4d IE |
862 | |
863 | } | |
f78e0e2e SY |
864 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
865 | ||
866 | int kvm_set_memory_region(struct kvm *kvm, | |
867 | struct kvm_userspace_memory_region *mem, | |
868 | int user_alloc) | |
869 | { | |
870 | int r; | |
871 | ||
79fac95e | 872 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 873 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 874 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
875 | return r; |
876 | } | |
210c7c4d IE |
877 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
878 | ||
1fe779f8 CO |
879 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
880 | struct | |
881 | kvm_userspace_memory_region *mem, | |
882 | int user_alloc) | |
210c7c4d | 883 | { |
e0d62c7f IE |
884 | if (mem->slot >= KVM_MEMORY_SLOTS) |
885 | return -EINVAL; | |
210c7c4d | 886 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
887 | } |
888 | ||
5bb064dc ZX |
889 | int kvm_get_dirty_log(struct kvm *kvm, |
890 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
891 | { |
892 | struct kvm_memory_slot *memslot; | |
893 | int r, i; | |
87bf6e7d | 894 | unsigned long n; |
6aa8b732 AK |
895 | unsigned long any = 0; |
896 | ||
6aa8b732 AK |
897 | r = -EINVAL; |
898 | if (log->slot >= KVM_MEMORY_SLOTS) | |
899 | goto out; | |
900 | ||
28a37544 | 901 | memslot = id_to_memslot(kvm->memslots, log->slot); |
6aa8b732 AK |
902 | r = -ENOENT; |
903 | if (!memslot->dirty_bitmap) | |
904 | goto out; | |
905 | ||
87bf6e7d | 906 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 907 | |
cd1a4a98 | 908 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
909 | any = memslot->dirty_bitmap[i]; |
910 | ||
911 | r = -EFAULT; | |
912 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
913 | goto out; | |
914 | ||
5bb064dc ZX |
915 | if (any) |
916 | *is_dirty = 1; | |
6aa8b732 AK |
917 | |
918 | r = 0; | |
6aa8b732 | 919 | out: |
6aa8b732 AK |
920 | return r; |
921 | } | |
922 | ||
db3fe4eb TY |
923 | bool kvm_largepages_enabled(void) |
924 | { | |
925 | return largepages_enabled; | |
926 | } | |
927 | ||
54dee993 MT |
928 | void kvm_disable_largepages(void) |
929 | { | |
930 | largepages_enabled = false; | |
931 | } | |
932 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
933 | ||
cea7bb21 IE |
934 | int is_error_page(struct page *page) |
935 | { | |
a2766325 | 936 | return IS_ERR(page); |
cea7bb21 IE |
937 | } |
938 | EXPORT_SYMBOL_GPL(is_error_page); | |
939 | ||
35149e21 AL |
940 | int is_error_pfn(pfn_t pfn) |
941 | { | |
a2766325 | 942 | return IS_ERR_VALUE(pfn); |
35149e21 AL |
943 | } |
944 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
945 | ||
a2766325 XG |
946 | static pfn_t get_bad_pfn(void) |
947 | { | |
948 | return -ENOENT; | |
949 | } | |
950 | ||
951 | pfn_t get_fault_pfn(void) | |
952 | { | |
953 | return -EFAULT; | |
954 | } | |
955 | EXPORT_SYMBOL_GPL(get_fault_pfn); | |
956 | ||
957 | static pfn_t get_hwpoison_pfn(void) | |
958 | { | |
959 | return -EHWPOISON; | |
960 | } | |
961 | ||
bf998156 HY |
962 | int is_hwpoison_pfn(pfn_t pfn) |
963 | { | |
a2766325 | 964 | return pfn == -EHWPOISON; |
bf998156 HY |
965 | } |
966 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
967 | ||
fce92dce XG |
968 | int is_noslot_pfn(pfn_t pfn) |
969 | { | |
a2766325 | 970 | return pfn == -ENOENT; |
fce92dce XG |
971 | } |
972 | EXPORT_SYMBOL_GPL(is_noslot_pfn); | |
973 | ||
974 | int is_invalid_pfn(pfn_t pfn) | |
975 | { | |
a2766325 | 976 | return !is_noslot_pfn(pfn) && is_error_pfn(pfn); |
fce92dce XG |
977 | } |
978 | EXPORT_SYMBOL_GPL(is_invalid_pfn); | |
979 | ||
a2766325 XG |
980 | struct page *get_bad_page(void) |
981 | { | |
982 | return ERR_PTR(-ENOENT); | |
983 | } | |
984 | ||
f9d46eb0 IE |
985 | static inline unsigned long bad_hva(void) |
986 | { | |
987 | return PAGE_OFFSET; | |
988 | } | |
989 | ||
990 | int kvm_is_error_hva(unsigned long addr) | |
991 | { | |
992 | return addr == bad_hva(); | |
993 | } | |
994 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
995 | ||
49c7754c GN |
996 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) |
997 | { | |
998 | return __gfn_to_memslot(kvm_memslots(kvm), gfn); | |
999 | } | |
a1f4d395 | 1000 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 1001 | |
e0d62c7f IE |
1002 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
1003 | { | |
bf3e05bc | 1004 | struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn); |
e0d62c7f | 1005 | |
bf3e05bc XG |
1006 | if (!memslot || memslot->id >= KVM_MEMORY_SLOTS || |
1007 | memslot->flags & KVM_MEMSLOT_INVALID) | |
1008 | return 0; | |
e0d62c7f | 1009 | |
bf3e05bc | 1010 | return 1; |
e0d62c7f IE |
1011 | } |
1012 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
1013 | ||
8f0b1ab6 JR |
1014 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
1015 | { | |
1016 | struct vm_area_struct *vma; | |
1017 | unsigned long addr, size; | |
1018 | ||
1019 | size = PAGE_SIZE; | |
1020 | ||
1021 | addr = gfn_to_hva(kvm, gfn); | |
1022 | if (kvm_is_error_hva(addr)) | |
1023 | return PAGE_SIZE; | |
1024 | ||
1025 | down_read(¤t->mm->mmap_sem); | |
1026 | vma = find_vma(current->mm, addr); | |
1027 | if (!vma) | |
1028 | goto out; | |
1029 | ||
1030 | size = vma_kernel_pagesize(vma); | |
1031 | ||
1032 | out: | |
1033 | up_read(¤t->mm->mmap_sem); | |
1034 | ||
1035 | return size; | |
1036 | } | |
1037 | ||
49c7754c | 1038 | static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn, |
48987781 | 1039 | gfn_t *nr_pages) |
539cb660 | 1040 | { |
bc6678a3 | 1041 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 1042 | return bad_hva(); |
48987781 XG |
1043 | |
1044 | if (nr_pages) | |
1045 | *nr_pages = slot->npages - (gfn - slot->base_gfn); | |
1046 | ||
f5c98031 | 1047 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 1048 | } |
48987781 XG |
1049 | |
1050 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) | |
1051 | { | |
49c7754c | 1052 | return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL); |
48987781 | 1053 | } |
0d150298 | 1054 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1055 | |
0857b9e9 GN |
1056 | int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm, |
1057 | unsigned long start, int write, struct page **page) | |
1058 | { | |
1059 | int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET; | |
1060 | ||
1061 | if (write) | |
1062 | flags |= FOLL_WRITE; | |
1063 | ||
1064 | return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL); | |
1065 | } | |
1066 | ||
fafc3dba HY |
1067 | static inline int check_user_page_hwpoison(unsigned long addr) |
1068 | { | |
1069 | int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE; | |
1070 | ||
1071 | rc = __get_user_pages(current, current->mm, addr, 1, | |
1072 | flags, NULL, NULL, NULL); | |
1073 | return rc == -EHWPOISON; | |
1074 | } | |
1075 | ||
d5661048 XG |
1076 | static pfn_t hva_to_pfn(unsigned long addr, bool atomic, bool *async, |
1077 | bool write_fault, bool *writable) | |
954bbbc2 | 1078 | { |
8d4e1288 | 1079 | struct page *page[1]; |
af585b92 | 1080 | int npages = 0; |
2e2e3738 | 1081 | pfn_t pfn; |
954bbbc2 | 1082 | |
af585b92 GN |
1083 | /* we can do it either atomically or asynchronously, not both */ |
1084 | BUG_ON(atomic && async); | |
1085 | ||
612819c3 MT |
1086 | BUG_ON(!write_fault && !writable); |
1087 | ||
1088 | if (writable) | |
1089 | *writable = true; | |
1090 | ||
af585b92 | 1091 | if (atomic || async) |
887c08ac | 1092 | npages = __get_user_pages_fast(addr, 1, 1, page); |
af585b92 GN |
1093 | |
1094 | if (unlikely(npages != 1) && !atomic) { | |
887c08ac | 1095 | might_sleep(); |
612819c3 MT |
1096 | |
1097 | if (writable) | |
1098 | *writable = write_fault; | |
1099 | ||
0857b9e9 GN |
1100 | if (async) { |
1101 | down_read(¤t->mm->mmap_sem); | |
1102 | npages = get_user_page_nowait(current, current->mm, | |
1103 | addr, write_fault, page); | |
1104 | up_read(¤t->mm->mmap_sem); | |
1105 | } else | |
1106 | npages = get_user_pages_fast(addr, 1, write_fault, | |
1107 | page); | |
612819c3 MT |
1108 | |
1109 | /* map read fault as writable if possible */ | |
1110 | if (unlikely(!write_fault) && npages == 1) { | |
1111 | struct page *wpage[1]; | |
1112 | ||
1113 | npages = __get_user_pages_fast(addr, 1, 1, wpage); | |
1114 | if (npages == 1) { | |
1115 | *writable = true; | |
1116 | put_page(page[0]); | |
1117 | page[0] = wpage[0]; | |
1118 | } | |
1119 | npages = 1; | |
1120 | } | |
887c08ac | 1121 | } |
539cb660 | 1122 | |
2e2e3738 AL |
1123 | if (unlikely(npages != 1)) { |
1124 | struct vm_area_struct *vma; | |
1125 | ||
887c08ac | 1126 | if (atomic) |
8030089f | 1127 | return get_fault_pfn(); |
887c08ac | 1128 | |
bbeb3406 | 1129 | down_read(¤t->mm->mmap_sem); |
0857b9e9 GN |
1130 | if (npages == -EHWPOISON || |
1131 | (!async && check_user_page_hwpoison(addr))) { | |
bbeb3406 | 1132 | up_read(¤t->mm->mmap_sem); |
a2766325 | 1133 | return get_hwpoison_pfn(); |
bf998156 HY |
1134 | } |
1135 | ||
8030089f | 1136 | vma = find_vma_intersection(current->mm, addr, addr+1); |
4c2155ce | 1137 | |
8030089f GN |
1138 | if (vma == NULL) |
1139 | pfn = get_fault_pfn(); | |
1140 | else if ((vma->vm_flags & VM_PFNMAP)) { | |
1141 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + | |
1142 | vma->vm_pgoff; | |
1143 | BUG_ON(!kvm_is_mmio_pfn(pfn)); | |
1144 | } else { | |
1145 | if (async && (vma->vm_flags & VM_WRITE)) | |
af585b92 | 1146 | *async = true; |
8030089f | 1147 | pfn = get_fault_pfn(); |
2e2e3738 | 1148 | } |
4c2155ce | 1149 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1150 | } else |
1151 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1152 | |
2e2e3738 | 1153 | return pfn; |
35149e21 AL |
1154 | } |
1155 | ||
d5661048 | 1156 | pfn_t hva_to_pfn_atomic(unsigned long addr) |
887c08ac | 1157 | { |
d5661048 | 1158 | return hva_to_pfn(addr, true, NULL, true, NULL); |
887c08ac XG |
1159 | } |
1160 | EXPORT_SYMBOL_GPL(hva_to_pfn_atomic); | |
1161 | ||
612819c3 MT |
1162 | static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async, |
1163 | bool write_fault, bool *writable) | |
506f0d6f MT |
1164 | { |
1165 | unsigned long addr; | |
1166 | ||
af585b92 GN |
1167 | if (async) |
1168 | *async = false; | |
1169 | ||
506f0d6f | 1170 | addr = gfn_to_hva(kvm, gfn); |
a2766325 XG |
1171 | if (kvm_is_error_hva(addr)) |
1172 | return get_bad_pfn(); | |
506f0d6f | 1173 | |
d5661048 | 1174 | return hva_to_pfn(addr, atomic, async, write_fault, writable); |
365fb3fd XG |
1175 | } |
1176 | ||
1177 | pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn) | |
1178 | { | |
612819c3 | 1179 | return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL); |
365fb3fd XG |
1180 | } |
1181 | EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic); | |
1182 | ||
612819c3 MT |
1183 | pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async, |
1184 | bool write_fault, bool *writable) | |
af585b92 | 1185 | { |
612819c3 | 1186 | return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable); |
af585b92 GN |
1187 | } |
1188 | EXPORT_SYMBOL_GPL(gfn_to_pfn_async); | |
1189 | ||
365fb3fd XG |
1190 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
1191 | { | |
612819c3 | 1192 | return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL); |
506f0d6f | 1193 | } |
35149e21 AL |
1194 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
1195 | ||
612819c3 MT |
1196 | pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, |
1197 | bool *writable) | |
1198 | { | |
1199 | return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable); | |
1200 | } | |
1201 | EXPORT_SYMBOL_GPL(gfn_to_pfn_prot); | |
1202 | ||
d5661048 | 1203 | pfn_t gfn_to_pfn_memslot(struct kvm_memory_slot *slot, gfn_t gfn) |
506f0d6f MT |
1204 | { |
1205 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
d5661048 | 1206 | return hva_to_pfn(addr, false, NULL, true, NULL); |
506f0d6f MT |
1207 | } |
1208 | ||
48987781 XG |
1209 | int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages, |
1210 | int nr_pages) | |
1211 | { | |
1212 | unsigned long addr; | |
1213 | gfn_t entry; | |
1214 | ||
49c7754c | 1215 | addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry); |
48987781 XG |
1216 | if (kvm_is_error_hva(addr)) |
1217 | return -1; | |
1218 | ||
1219 | if (entry < nr_pages) | |
1220 | return 0; | |
1221 | ||
1222 | return __get_user_pages_fast(addr, nr_pages, 1, pages); | |
1223 | } | |
1224 | EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic); | |
1225 | ||
a2766325 XG |
1226 | static struct page *kvm_pfn_to_page(pfn_t pfn) |
1227 | { | |
1228 | WARN_ON(kvm_is_mmio_pfn(pfn)); | |
1229 | ||
1230 | if (is_error_pfn(pfn) || kvm_is_mmio_pfn(pfn)) | |
1231 | return get_bad_page(); | |
1232 | ||
1233 | return pfn_to_page(pfn); | |
1234 | } | |
1235 | ||
35149e21 AL |
1236 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1237 | { | |
2e2e3738 AL |
1238 | pfn_t pfn; |
1239 | ||
1240 | pfn = gfn_to_pfn(kvm, gfn); | |
2e2e3738 | 1241 | |
a2766325 | 1242 | return kvm_pfn_to_page(pfn); |
954bbbc2 | 1243 | } |
aab61cc0 | 1244 | |
954bbbc2 AK |
1245 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1246 | ||
b4231d61 IE |
1247 | void kvm_release_page_clean(struct page *page) |
1248 | { | |
a2766325 XG |
1249 | if (!is_error_page(page)) |
1250 | kvm_release_pfn_clean(page_to_pfn(page)); | |
b4231d61 IE |
1251 | } |
1252 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1253 | ||
35149e21 AL |
1254 | void kvm_release_pfn_clean(pfn_t pfn) |
1255 | { | |
a2766325 | 1256 | if (!is_error_pfn(pfn) && !kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1257 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1258 | } |
1259 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1260 | ||
b4231d61 | 1261 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1262 | { |
a2766325 XG |
1263 | WARN_ON(is_error_page(page)); |
1264 | ||
35149e21 AL |
1265 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1266 | } | |
1267 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1268 | ||
1269 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1270 | { | |
1271 | kvm_set_pfn_dirty(pfn); | |
1272 | kvm_release_pfn_clean(pfn); | |
1273 | } | |
1274 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1275 | ||
1276 | void kvm_set_page_dirty(struct page *page) | |
1277 | { | |
1278 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1279 | } | |
1280 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1281 | ||
1282 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1283 | { | |
c77fb9dc | 1284 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1285 | struct page *page = pfn_to_page(pfn); |
1286 | if (!PageReserved(page)) | |
1287 | SetPageDirty(page); | |
1288 | } | |
8a7ae055 | 1289 | } |
35149e21 AL |
1290 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1291 | ||
1292 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1293 | { | |
c77fb9dc | 1294 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1295 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1296 | } |
1297 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1298 | ||
1299 | void kvm_get_pfn(pfn_t pfn) | |
1300 | { | |
c77fb9dc | 1301 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1302 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1303 | } |
1304 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1305 | |
195aefde IE |
1306 | static int next_segment(unsigned long len, int offset) |
1307 | { | |
1308 | if (len > PAGE_SIZE - offset) | |
1309 | return PAGE_SIZE - offset; | |
1310 | else | |
1311 | return len; | |
1312 | } | |
1313 | ||
1314 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1315 | int len) | |
1316 | { | |
e0506bcb IE |
1317 | int r; |
1318 | unsigned long addr; | |
195aefde | 1319 | |
e0506bcb IE |
1320 | addr = gfn_to_hva(kvm, gfn); |
1321 | if (kvm_is_error_hva(addr)) | |
1322 | return -EFAULT; | |
fa3d315a | 1323 | r = __copy_from_user(data, (void __user *)addr + offset, len); |
e0506bcb | 1324 | if (r) |
195aefde | 1325 | return -EFAULT; |
195aefde IE |
1326 | return 0; |
1327 | } | |
1328 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1329 | ||
1330 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1331 | { | |
1332 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1333 | int seg; | |
1334 | int offset = offset_in_page(gpa); | |
1335 | int ret; | |
1336 | ||
1337 | while ((seg = next_segment(len, offset)) != 0) { | |
1338 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1339 | if (ret < 0) | |
1340 | return ret; | |
1341 | offset = 0; | |
1342 | len -= seg; | |
1343 | data += seg; | |
1344 | ++gfn; | |
1345 | } | |
1346 | return 0; | |
1347 | } | |
1348 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1349 | ||
7ec54588 MT |
1350 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1351 | unsigned long len) | |
1352 | { | |
1353 | int r; | |
1354 | unsigned long addr; | |
1355 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1356 | int offset = offset_in_page(gpa); | |
1357 | ||
1358 | addr = gfn_to_hva(kvm, gfn); | |
1359 | if (kvm_is_error_hva(addr)) | |
1360 | return -EFAULT; | |
0aac03f0 | 1361 | pagefault_disable(); |
7ec54588 | 1362 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1363 | pagefault_enable(); |
7ec54588 MT |
1364 | if (r) |
1365 | return -EFAULT; | |
1366 | return 0; | |
1367 | } | |
1368 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1369 | ||
195aefde IE |
1370 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1371 | int offset, int len) | |
1372 | { | |
e0506bcb IE |
1373 | int r; |
1374 | unsigned long addr; | |
195aefde | 1375 | |
e0506bcb IE |
1376 | addr = gfn_to_hva(kvm, gfn); |
1377 | if (kvm_is_error_hva(addr)) | |
1378 | return -EFAULT; | |
8b0cedff | 1379 | r = __copy_to_user((void __user *)addr + offset, data, len); |
e0506bcb | 1380 | if (r) |
195aefde | 1381 | return -EFAULT; |
195aefde IE |
1382 | mark_page_dirty(kvm, gfn); |
1383 | return 0; | |
1384 | } | |
1385 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1386 | ||
1387 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1388 | unsigned long len) | |
1389 | { | |
1390 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1391 | int seg; | |
1392 | int offset = offset_in_page(gpa); | |
1393 | int ret; | |
1394 | ||
1395 | while ((seg = next_segment(len, offset)) != 0) { | |
1396 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1397 | if (ret < 0) | |
1398 | return ret; | |
1399 | offset = 0; | |
1400 | len -= seg; | |
1401 | data += seg; | |
1402 | ++gfn; | |
1403 | } | |
1404 | return 0; | |
1405 | } | |
1406 | ||
49c7754c GN |
1407 | int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1408 | gpa_t gpa) | |
1409 | { | |
1410 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1411 | int offset = offset_in_page(gpa); | |
1412 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1413 | ||
1414 | ghc->gpa = gpa; | |
1415 | ghc->generation = slots->generation; | |
9d4cba7f | 1416 | ghc->memslot = gfn_to_memslot(kvm, gfn); |
49c7754c GN |
1417 | ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL); |
1418 | if (!kvm_is_error_hva(ghc->hva)) | |
1419 | ghc->hva += offset; | |
1420 | else | |
1421 | return -EFAULT; | |
1422 | ||
1423 | return 0; | |
1424 | } | |
1425 | EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init); | |
1426 | ||
1427 | int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, | |
1428 | void *data, unsigned long len) | |
1429 | { | |
1430 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1431 | int r; | |
1432 | ||
1433 | if (slots->generation != ghc->generation) | |
1434 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1435 | ||
1436 | if (kvm_is_error_hva(ghc->hva)) | |
1437 | return -EFAULT; | |
1438 | ||
8b0cedff | 1439 | r = __copy_to_user((void __user *)ghc->hva, data, len); |
49c7754c GN |
1440 | if (r) |
1441 | return -EFAULT; | |
1442 | mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT); | |
1443 | ||
1444 | return 0; | |
1445 | } | |
1446 | EXPORT_SYMBOL_GPL(kvm_write_guest_cached); | |
1447 | ||
e03b644f GN |
1448 | int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1449 | void *data, unsigned long len) | |
1450 | { | |
1451 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1452 | int r; | |
1453 | ||
1454 | if (slots->generation != ghc->generation) | |
1455 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1456 | ||
1457 | if (kvm_is_error_hva(ghc->hva)) | |
1458 | return -EFAULT; | |
1459 | ||
1460 | r = __copy_from_user(data, (void __user *)ghc->hva, len); | |
1461 | if (r) | |
1462 | return -EFAULT; | |
1463 | ||
1464 | return 0; | |
1465 | } | |
1466 | EXPORT_SYMBOL_GPL(kvm_read_guest_cached); | |
1467 | ||
195aefde IE |
1468 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) |
1469 | { | |
3bcc8a8c HC |
1470 | return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page, |
1471 | offset, len); | |
195aefde IE |
1472 | } |
1473 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1474 | ||
1475 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1476 | { | |
1477 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1478 | int seg; | |
1479 | int offset = offset_in_page(gpa); | |
1480 | int ret; | |
1481 | ||
1482 | while ((seg = next_segment(len, offset)) != 0) { | |
1483 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1484 | if (ret < 0) | |
1485 | return ret; | |
1486 | offset = 0; | |
1487 | len -= seg; | |
1488 | ++gfn; | |
1489 | } | |
1490 | return 0; | |
1491 | } | |
1492 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1493 | ||
49c7754c GN |
1494 | void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, |
1495 | gfn_t gfn) | |
6aa8b732 | 1496 | { |
7e9d619d RR |
1497 | if (memslot && memslot->dirty_bitmap) { |
1498 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1499 | |
93474b25 TY |
1500 | /* TODO: introduce set_bit_le() and use it */ |
1501 | test_and_set_bit_le(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1502 | } |
1503 | } | |
1504 | ||
49c7754c GN |
1505 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1506 | { | |
1507 | struct kvm_memory_slot *memslot; | |
1508 | ||
1509 | memslot = gfn_to_memslot(kvm, gfn); | |
1510 | mark_page_dirty_in_slot(kvm, memslot, gfn); | |
1511 | } | |
1512 | ||
b6958ce4 ED |
1513 | /* |
1514 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1515 | */ | |
8776e519 | 1516 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1517 | { |
e5c239cf MT |
1518 | DEFINE_WAIT(wait); |
1519 | ||
1520 | for (;;) { | |
1521 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1522 | ||
a1b37100 | 1523 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1524 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1525 | break; |
d7690175 | 1526 | } |
09cec754 GN |
1527 | if (kvm_cpu_has_pending_timer(vcpu)) |
1528 | break; | |
e5c239cf MT |
1529 | if (signal_pending(current)) |
1530 | break; | |
1531 | ||
b6958ce4 | 1532 | schedule(); |
b6958ce4 | 1533 | } |
d3bef15f | 1534 | |
e5c239cf | 1535 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1536 | } |
1537 | ||
8c84780d | 1538 | #ifndef CONFIG_S390 |
b6d33834 CD |
1539 | /* |
1540 | * Kick a sleeping VCPU, or a guest VCPU in guest mode, into host kernel mode. | |
1541 | */ | |
1542 | void kvm_vcpu_kick(struct kvm_vcpu *vcpu) | |
1543 | { | |
1544 | int me; | |
1545 | int cpu = vcpu->cpu; | |
1546 | wait_queue_head_t *wqp; | |
1547 | ||
1548 | wqp = kvm_arch_vcpu_wq(vcpu); | |
1549 | if (waitqueue_active(wqp)) { | |
1550 | wake_up_interruptible(wqp); | |
1551 | ++vcpu->stat.halt_wakeup; | |
1552 | } | |
1553 | ||
1554 | me = get_cpu(); | |
1555 | if (cpu != me && (unsigned)cpu < nr_cpu_ids && cpu_online(cpu)) | |
1556 | if (kvm_arch_vcpu_should_kick(vcpu)) | |
1557 | smp_send_reschedule(cpu); | |
1558 | put_cpu(); | |
1559 | } | |
8c84780d | 1560 | #endif /* !CONFIG_S390 */ |
b6d33834 | 1561 | |
6aa8b732 AK |
1562 | void kvm_resched(struct kvm_vcpu *vcpu) |
1563 | { | |
3fca0365 YD |
1564 | if (!need_resched()) |
1565 | return; | |
6aa8b732 | 1566 | cond_resched(); |
6aa8b732 AK |
1567 | } |
1568 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1569 | ||
41628d33 KW |
1570 | bool kvm_vcpu_yield_to(struct kvm_vcpu *target) |
1571 | { | |
1572 | struct pid *pid; | |
1573 | struct task_struct *task = NULL; | |
1574 | ||
1575 | rcu_read_lock(); | |
1576 | pid = rcu_dereference(target->pid); | |
1577 | if (pid) | |
1578 | task = get_pid_task(target->pid, PIDTYPE_PID); | |
1579 | rcu_read_unlock(); | |
1580 | if (!task) | |
1581 | return false; | |
1582 | if (task->flags & PF_VCPU) { | |
1583 | put_task_struct(task); | |
1584 | return false; | |
1585 | } | |
1586 | if (yield_to(task, 1)) { | |
1587 | put_task_struct(task); | |
1588 | return true; | |
1589 | } | |
1590 | put_task_struct(task); | |
1591 | return false; | |
1592 | } | |
1593 | EXPORT_SYMBOL_GPL(kvm_vcpu_yield_to); | |
1594 | ||
06e48c51 R |
1595 | #ifdef CONFIG_HAVE_KVM_CPU_RELAX_INTERCEPT |
1596 | /* | |
1597 | * Helper that checks whether a VCPU is eligible for directed yield. | |
1598 | * Most eligible candidate to yield is decided by following heuristics: | |
1599 | * | |
1600 | * (a) VCPU which has not done pl-exit or cpu relax intercepted recently | |
1601 | * (preempted lock holder), indicated by @in_spin_loop. | |
1602 | * Set at the beiginning and cleared at the end of interception/PLE handler. | |
1603 | * | |
1604 | * (b) VCPU which has done pl-exit/ cpu relax intercepted but did not get | |
1605 | * chance last time (mostly it has become eligible now since we have probably | |
1606 | * yielded to lockholder in last iteration. This is done by toggling | |
1607 | * @dy_eligible each time a VCPU checked for eligibility.) | |
1608 | * | |
1609 | * Yielding to a recently pl-exited/cpu relax intercepted VCPU before yielding | |
1610 | * to preempted lock-holder could result in wrong VCPU selection and CPU | |
1611 | * burning. Giving priority for a potential lock-holder increases lock | |
1612 | * progress. | |
1613 | * | |
1614 | * Since algorithm is based on heuristics, accessing another VCPU data without | |
1615 | * locking does not harm. It may result in trying to yield to same VCPU, fail | |
1616 | * and continue with next VCPU and so on. | |
1617 | */ | |
1618 | bool kvm_vcpu_eligible_for_directed_yield(struct kvm_vcpu *vcpu) | |
1619 | { | |
1620 | bool eligible; | |
1621 | ||
1622 | eligible = !vcpu->spin_loop.in_spin_loop || | |
1623 | (vcpu->spin_loop.in_spin_loop && | |
1624 | vcpu->spin_loop.dy_eligible); | |
1625 | ||
1626 | if (vcpu->spin_loop.in_spin_loop) | |
1627 | kvm_vcpu_set_dy_eligible(vcpu, !vcpu->spin_loop.dy_eligible); | |
1628 | ||
1629 | return eligible; | |
1630 | } | |
1631 | #endif | |
217ece61 | 1632 | void kvm_vcpu_on_spin(struct kvm_vcpu *me) |
d255f4f2 | 1633 | { |
217ece61 RR |
1634 | struct kvm *kvm = me->kvm; |
1635 | struct kvm_vcpu *vcpu; | |
1636 | int last_boosted_vcpu = me->kvm->last_boosted_vcpu; | |
1637 | int yielded = 0; | |
1638 | int pass; | |
1639 | int i; | |
d255f4f2 | 1640 | |
4c088493 | 1641 | kvm_vcpu_set_in_spin_loop(me, true); |
217ece61 RR |
1642 | /* |
1643 | * We boost the priority of a VCPU that is runnable but not | |
1644 | * currently running, because it got preempted by something | |
1645 | * else and called schedule in __vcpu_run. Hopefully that | |
1646 | * VCPU is holding the lock that we need and will release it. | |
1647 | * We approximate round-robin by starting at the last boosted VCPU. | |
1648 | */ | |
1649 | for (pass = 0; pass < 2 && !yielded; pass++) { | |
1650 | kvm_for_each_vcpu(i, vcpu, kvm) { | |
5cfc2aab | 1651 | if (!pass && i <= last_boosted_vcpu) { |
217ece61 RR |
1652 | i = last_boosted_vcpu; |
1653 | continue; | |
1654 | } else if (pass && i > last_boosted_vcpu) | |
1655 | break; | |
1656 | if (vcpu == me) | |
1657 | continue; | |
1658 | if (waitqueue_active(&vcpu->wq)) | |
1659 | continue; | |
06e48c51 R |
1660 | if (!kvm_vcpu_eligible_for_directed_yield(vcpu)) |
1661 | continue; | |
41628d33 | 1662 | if (kvm_vcpu_yield_to(vcpu)) { |
217ece61 RR |
1663 | kvm->last_boosted_vcpu = i; |
1664 | yielded = 1; | |
1665 | break; | |
1666 | } | |
217ece61 RR |
1667 | } |
1668 | } | |
4c088493 | 1669 | kvm_vcpu_set_in_spin_loop(me, false); |
06e48c51 R |
1670 | |
1671 | /* Ensure vcpu is not eligible during next spinloop */ | |
1672 | kvm_vcpu_set_dy_eligible(me, false); | |
d255f4f2 ZE |
1673 | } |
1674 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1675 | ||
e4a533a4 | 1676 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1677 | { |
1678 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1679 | struct page *page; |
1680 | ||
e4a533a4 | 1681 | if (vmf->pgoff == 0) |
039576c0 | 1682 | page = virt_to_page(vcpu->run); |
09566765 | 1683 | #ifdef CONFIG_X86 |
e4a533a4 | 1684 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1685 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1686 | #endif |
1687 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1688 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1689 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1690 | #endif |
039576c0 | 1691 | else |
5b1c1493 | 1692 | return kvm_arch_vcpu_fault(vcpu, vmf); |
9a2bb7f4 | 1693 | get_page(page); |
e4a533a4 | 1694 | vmf->page = page; |
1695 | return 0; | |
9a2bb7f4 AK |
1696 | } |
1697 | ||
f0f37e2f | 1698 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1699 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1700 | }; |
1701 | ||
1702 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1703 | { | |
1704 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1705 | return 0; | |
1706 | } | |
1707 | ||
bccf2150 AK |
1708 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1709 | { | |
1710 | struct kvm_vcpu *vcpu = filp->private_data; | |
1711 | ||
66c0b394 | 1712 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1713 | return 0; |
1714 | } | |
1715 | ||
3d3aab1b | 1716 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1717 | .release = kvm_vcpu_release, |
1718 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1dda606c AG |
1719 | #ifdef CONFIG_COMPAT |
1720 | .compat_ioctl = kvm_vcpu_compat_ioctl, | |
1721 | #endif | |
9a2bb7f4 | 1722 | .mmap = kvm_vcpu_mmap, |
6038f373 | 1723 | .llseek = noop_llseek, |
bccf2150 AK |
1724 | }; |
1725 | ||
1726 | /* | |
1727 | * Allocates an inode for the vcpu. | |
1728 | */ | |
1729 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1730 | { | |
628ff7c1 | 1731 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1732 | } |
1733 | ||
c5ea7660 AK |
1734 | /* |
1735 | * Creates some virtual cpus. Good luck creating more than one. | |
1736 | */ | |
73880c80 | 1737 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1738 | { |
1739 | int r; | |
988a2cae | 1740 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1741 | |
73880c80 | 1742 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1743 | if (IS_ERR(vcpu)) |
1744 | return PTR_ERR(vcpu); | |
c5ea7660 | 1745 | |
15ad7146 AK |
1746 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1747 | ||
26e5215f AK |
1748 | r = kvm_arch_vcpu_setup(vcpu); |
1749 | if (r) | |
d780592b | 1750 | goto vcpu_destroy; |
26e5215f | 1751 | |
11ec2804 | 1752 | mutex_lock(&kvm->lock); |
3e515705 AK |
1753 | if (!kvm_vcpu_compatible(vcpu)) { |
1754 | r = -EINVAL; | |
1755 | goto unlock_vcpu_destroy; | |
1756 | } | |
73880c80 GN |
1757 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1758 | r = -EINVAL; | |
d780592b | 1759 | goto unlock_vcpu_destroy; |
fb3f0f51 | 1760 | } |
73880c80 | 1761 | |
988a2cae GN |
1762 | kvm_for_each_vcpu(r, v, kvm) |
1763 | if (v->vcpu_id == id) { | |
73880c80 | 1764 | r = -EEXIST; |
d780592b | 1765 | goto unlock_vcpu_destroy; |
73880c80 GN |
1766 | } |
1767 | ||
1768 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1769 | |
fb3f0f51 | 1770 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1771 | kvm_get_kvm(kvm); |
bccf2150 | 1772 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1773 | if (r < 0) { |
1774 | kvm_put_kvm(kvm); | |
d780592b | 1775 | goto unlock_vcpu_destroy; |
73880c80 GN |
1776 | } |
1777 | ||
1778 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1779 | smp_wmb(); | |
1780 | atomic_inc(&kvm->online_vcpus); | |
1781 | ||
73880c80 | 1782 | mutex_unlock(&kvm->lock); |
fb3f0f51 | 1783 | return r; |
39c3b86e | 1784 | |
d780592b | 1785 | unlock_vcpu_destroy: |
7d8fece6 | 1786 | mutex_unlock(&kvm->lock); |
d780592b | 1787 | vcpu_destroy: |
d40ccc62 | 1788 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1789 | return r; |
1790 | } | |
1791 | ||
1961d276 AK |
1792 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1793 | { | |
1794 | if (sigset) { | |
1795 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1796 | vcpu->sigset_active = 1; | |
1797 | vcpu->sigset = *sigset; | |
1798 | } else | |
1799 | vcpu->sigset_active = 0; | |
1800 | return 0; | |
1801 | } | |
1802 | ||
bccf2150 AK |
1803 | static long kvm_vcpu_ioctl(struct file *filp, |
1804 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1805 | { |
bccf2150 | 1806 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1807 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1808 | int r; |
fa3795a7 DH |
1809 | struct kvm_fpu *fpu = NULL; |
1810 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1811 | |
6d4e4c4f AK |
1812 | if (vcpu->kvm->mm != current->mm) |
1813 | return -EIO; | |
2122ff5e AK |
1814 | |
1815 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1816 | /* | |
1817 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1818 | * so vcpu_load() would break it. | |
1819 | */ | |
1820 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1821 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1822 | #endif | |
1823 | ||
1824 | ||
1825 | vcpu_load(vcpu); | |
6aa8b732 | 1826 | switch (ioctl) { |
9a2bb7f4 | 1827 | case KVM_RUN: |
f0fe5108 AK |
1828 | r = -EINVAL; |
1829 | if (arg) | |
1830 | goto out; | |
b6c7a5dc | 1831 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
64be5007 | 1832 | trace_kvm_userspace_exit(vcpu->run->exit_reason, r); |
6aa8b732 | 1833 | break; |
6aa8b732 | 1834 | case KVM_GET_REGS: { |
3e4bb3ac | 1835 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1836 | |
3e4bb3ac XZ |
1837 | r = -ENOMEM; |
1838 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1839 | if (!kvm_regs) | |
6aa8b732 | 1840 | goto out; |
3e4bb3ac XZ |
1841 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1842 | if (r) | |
1843 | goto out_free1; | |
6aa8b732 | 1844 | r = -EFAULT; |
3e4bb3ac XZ |
1845 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1846 | goto out_free1; | |
6aa8b732 | 1847 | r = 0; |
3e4bb3ac XZ |
1848 | out_free1: |
1849 | kfree(kvm_regs); | |
6aa8b732 AK |
1850 | break; |
1851 | } | |
1852 | case KVM_SET_REGS: { | |
3e4bb3ac | 1853 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1854 | |
3e4bb3ac | 1855 | r = -ENOMEM; |
ff5c2c03 SL |
1856 | kvm_regs = memdup_user(argp, sizeof(*kvm_regs)); |
1857 | if (IS_ERR(kvm_regs)) { | |
1858 | r = PTR_ERR(kvm_regs); | |
6aa8b732 | 1859 | goto out; |
ff5c2c03 | 1860 | } |
3e4bb3ac | 1861 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); |
6aa8b732 | 1862 | if (r) |
3e4bb3ac | 1863 | goto out_free2; |
6aa8b732 | 1864 | r = 0; |
3e4bb3ac XZ |
1865 | out_free2: |
1866 | kfree(kvm_regs); | |
6aa8b732 AK |
1867 | break; |
1868 | } | |
1869 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1870 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1871 | r = -ENOMEM; | |
1872 | if (!kvm_sregs) | |
1873 | goto out; | |
1874 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1875 | if (r) |
1876 | goto out; | |
1877 | r = -EFAULT; | |
fa3795a7 | 1878 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1879 | goto out; |
1880 | r = 0; | |
1881 | break; | |
1882 | } | |
1883 | case KVM_SET_SREGS: { | |
ff5c2c03 SL |
1884 | kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs)); |
1885 | if (IS_ERR(kvm_sregs)) { | |
1886 | r = PTR_ERR(kvm_sregs); | |
6aa8b732 | 1887 | goto out; |
ff5c2c03 | 1888 | } |
fa3795a7 | 1889 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1890 | if (r) |
1891 | goto out; | |
1892 | r = 0; | |
1893 | break; | |
1894 | } | |
62d9f0db MT |
1895 | case KVM_GET_MP_STATE: { |
1896 | struct kvm_mp_state mp_state; | |
1897 | ||
1898 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1899 | if (r) | |
1900 | goto out; | |
1901 | r = -EFAULT; | |
1902 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1903 | goto out; | |
1904 | r = 0; | |
1905 | break; | |
1906 | } | |
1907 | case KVM_SET_MP_STATE: { | |
1908 | struct kvm_mp_state mp_state; | |
1909 | ||
1910 | r = -EFAULT; | |
1911 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1912 | goto out; | |
1913 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1914 | if (r) | |
1915 | goto out; | |
1916 | r = 0; | |
1917 | break; | |
1918 | } | |
6aa8b732 AK |
1919 | case KVM_TRANSLATE: { |
1920 | struct kvm_translation tr; | |
1921 | ||
1922 | r = -EFAULT; | |
2f366987 | 1923 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1924 | goto out; |
8b006791 | 1925 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1926 | if (r) |
1927 | goto out; | |
1928 | r = -EFAULT; | |
2f366987 | 1929 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1930 | goto out; |
1931 | r = 0; | |
1932 | break; | |
1933 | } | |
d0bfb940 JK |
1934 | case KVM_SET_GUEST_DEBUG: { |
1935 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1936 | |
1937 | r = -EFAULT; | |
2f366987 | 1938 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1939 | goto out; |
d0bfb940 | 1940 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1941 | if (r) |
1942 | goto out; | |
1943 | r = 0; | |
1944 | break; | |
1945 | } | |
1961d276 AK |
1946 | case KVM_SET_SIGNAL_MASK: { |
1947 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1948 | struct kvm_signal_mask kvm_sigmask; | |
1949 | sigset_t sigset, *p; | |
1950 | ||
1951 | p = NULL; | |
1952 | if (argp) { | |
1953 | r = -EFAULT; | |
1954 | if (copy_from_user(&kvm_sigmask, argp, | |
1955 | sizeof kvm_sigmask)) | |
1956 | goto out; | |
1957 | r = -EINVAL; | |
1958 | if (kvm_sigmask.len != sizeof sigset) | |
1959 | goto out; | |
1960 | r = -EFAULT; | |
1961 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1962 | sizeof sigset)) | |
1963 | goto out; | |
1964 | p = &sigset; | |
1965 | } | |
376d41ff | 1966 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1967 | break; |
1968 | } | |
b8836737 | 1969 | case KVM_GET_FPU: { |
fa3795a7 DH |
1970 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1971 | r = -ENOMEM; | |
1972 | if (!fpu) | |
1973 | goto out; | |
1974 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1975 | if (r) |
1976 | goto out; | |
1977 | r = -EFAULT; | |
fa3795a7 | 1978 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1979 | goto out; |
1980 | r = 0; | |
1981 | break; | |
1982 | } | |
1983 | case KVM_SET_FPU: { | |
ff5c2c03 SL |
1984 | fpu = memdup_user(argp, sizeof(*fpu)); |
1985 | if (IS_ERR(fpu)) { | |
1986 | r = PTR_ERR(fpu); | |
b8836737 | 1987 | goto out; |
ff5c2c03 | 1988 | } |
fa3795a7 | 1989 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1990 | if (r) |
1991 | goto out; | |
1992 | r = 0; | |
1993 | break; | |
1994 | } | |
bccf2150 | 1995 | default: |
313a3dc7 | 1996 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1997 | } |
1998 | out: | |
2122ff5e | 1999 | vcpu_put(vcpu); |
fa3795a7 DH |
2000 | kfree(fpu); |
2001 | kfree(kvm_sregs); | |
bccf2150 AK |
2002 | return r; |
2003 | } | |
2004 | ||
1dda606c AG |
2005 | #ifdef CONFIG_COMPAT |
2006 | static long kvm_vcpu_compat_ioctl(struct file *filp, | |
2007 | unsigned int ioctl, unsigned long arg) | |
2008 | { | |
2009 | struct kvm_vcpu *vcpu = filp->private_data; | |
2010 | void __user *argp = compat_ptr(arg); | |
2011 | int r; | |
2012 | ||
2013 | if (vcpu->kvm->mm != current->mm) | |
2014 | return -EIO; | |
2015 | ||
2016 | switch (ioctl) { | |
2017 | case KVM_SET_SIGNAL_MASK: { | |
2018 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
2019 | struct kvm_signal_mask kvm_sigmask; | |
2020 | compat_sigset_t csigset; | |
2021 | sigset_t sigset; | |
2022 | ||
2023 | if (argp) { | |
2024 | r = -EFAULT; | |
2025 | if (copy_from_user(&kvm_sigmask, argp, | |
2026 | sizeof kvm_sigmask)) | |
2027 | goto out; | |
2028 | r = -EINVAL; | |
2029 | if (kvm_sigmask.len != sizeof csigset) | |
2030 | goto out; | |
2031 | r = -EFAULT; | |
2032 | if (copy_from_user(&csigset, sigmask_arg->sigset, | |
2033 | sizeof csigset)) | |
2034 | goto out; | |
2035 | } | |
2036 | sigset_from_compat(&sigset, &csigset); | |
2037 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
2038 | break; | |
2039 | } | |
2040 | default: | |
2041 | r = kvm_vcpu_ioctl(filp, ioctl, arg); | |
2042 | } | |
2043 | ||
2044 | out: | |
2045 | return r; | |
2046 | } | |
2047 | #endif | |
2048 | ||
bccf2150 AK |
2049 | static long kvm_vm_ioctl(struct file *filp, |
2050 | unsigned int ioctl, unsigned long arg) | |
2051 | { | |
2052 | struct kvm *kvm = filp->private_data; | |
2053 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 2054 | int r; |
bccf2150 | 2055 | |
6d4e4c4f AK |
2056 | if (kvm->mm != current->mm) |
2057 | return -EIO; | |
bccf2150 AK |
2058 | switch (ioctl) { |
2059 | case KVM_CREATE_VCPU: | |
2060 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
2061 | if (r < 0) | |
2062 | goto out; | |
2063 | break; | |
6fc138d2 IE |
2064 | case KVM_SET_USER_MEMORY_REGION: { |
2065 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2066 | ||
2067 | r = -EFAULT; | |
2068 | if (copy_from_user(&kvm_userspace_mem, argp, | |
2069 | sizeof kvm_userspace_mem)) | |
2070 | goto out; | |
2071 | ||
2072 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
2073 | if (r) |
2074 | goto out; | |
2075 | break; | |
2076 | } | |
2077 | case KVM_GET_DIRTY_LOG: { | |
2078 | struct kvm_dirty_log log; | |
2079 | ||
2080 | r = -EFAULT; | |
2f366987 | 2081 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2082 | goto out; |
2c6f5df9 | 2083 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2084 | if (r) |
2085 | goto out; | |
2086 | break; | |
2087 | } | |
5f94c174 LV |
2088 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2089 | case KVM_REGISTER_COALESCED_MMIO: { | |
2090 | struct kvm_coalesced_mmio_zone zone; | |
2091 | r = -EFAULT; | |
2092 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2093 | goto out; | |
5f94c174 LV |
2094 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
2095 | if (r) | |
2096 | goto out; | |
2097 | r = 0; | |
2098 | break; | |
2099 | } | |
2100 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
2101 | struct kvm_coalesced_mmio_zone zone; | |
2102 | r = -EFAULT; | |
2103 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2104 | goto out; | |
5f94c174 LV |
2105 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
2106 | if (r) | |
2107 | goto out; | |
2108 | r = 0; | |
2109 | break; | |
2110 | } | |
2111 | #endif | |
721eecbf GH |
2112 | case KVM_IRQFD: { |
2113 | struct kvm_irqfd data; | |
2114 | ||
2115 | r = -EFAULT; | |
2116 | if (copy_from_user(&data, argp, sizeof data)) | |
2117 | goto out; | |
d4db2935 | 2118 | r = kvm_irqfd(kvm, &data); |
721eecbf GH |
2119 | break; |
2120 | } | |
d34e6b17 GH |
2121 | case KVM_IOEVENTFD: { |
2122 | struct kvm_ioeventfd data; | |
2123 | ||
2124 | r = -EFAULT; | |
2125 | if (copy_from_user(&data, argp, sizeof data)) | |
2126 | goto out; | |
2127 | r = kvm_ioeventfd(kvm, &data); | |
2128 | break; | |
2129 | } | |
73880c80 GN |
2130 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2131 | case KVM_SET_BOOT_CPU_ID: | |
2132 | r = 0; | |
894a9c55 | 2133 | mutex_lock(&kvm->lock); |
73880c80 GN |
2134 | if (atomic_read(&kvm->online_vcpus) != 0) |
2135 | r = -EBUSY; | |
2136 | else | |
2137 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 2138 | mutex_unlock(&kvm->lock); |
73880c80 | 2139 | break; |
07975ad3 JK |
2140 | #endif |
2141 | #ifdef CONFIG_HAVE_KVM_MSI | |
2142 | case KVM_SIGNAL_MSI: { | |
2143 | struct kvm_msi msi; | |
2144 | ||
2145 | r = -EFAULT; | |
2146 | if (copy_from_user(&msi, argp, sizeof msi)) | |
2147 | goto out; | |
2148 | r = kvm_send_userspace_msi(kvm, &msi); | |
2149 | break; | |
2150 | } | |
23d43cf9 CD |
2151 | #endif |
2152 | #ifdef __KVM_HAVE_IRQ_LINE | |
2153 | case KVM_IRQ_LINE_STATUS: | |
2154 | case KVM_IRQ_LINE: { | |
2155 | struct kvm_irq_level irq_event; | |
2156 | ||
2157 | r = -EFAULT; | |
2158 | if (copy_from_user(&irq_event, argp, sizeof irq_event)) | |
2159 | goto out; | |
2160 | ||
2161 | r = kvm_vm_ioctl_irq_line(kvm, &irq_event); | |
2162 | if (r) | |
2163 | goto out; | |
2164 | ||
2165 | r = -EFAULT; | |
2166 | if (ioctl == KVM_IRQ_LINE_STATUS) { | |
2167 | if (copy_to_user(argp, &irq_event, sizeof irq_event)) | |
2168 | goto out; | |
2169 | } | |
2170 | ||
2171 | r = 0; | |
2172 | break; | |
2173 | } | |
73880c80 | 2174 | #endif |
f17abe9a | 2175 | default: |
1fe779f8 | 2176 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
2177 | if (r == -ENOTTY) |
2178 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
2179 | } |
2180 | out: | |
2181 | return r; | |
2182 | } | |
2183 | ||
6ff5894c AB |
2184 | #ifdef CONFIG_COMPAT |
2185 | struct compat_kvm_dirty_log { | |
2186 | __u32 slot; | |
2187 | __u32 padding1; | |
2188 | union { | |
2189 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
2190 | __u64 padding2; | |
2191 | }; | |
2192 | }; | |
2193 | ||
2194 | static long kvm_vm_compat_ioctl(struct file *filp, | |
2195 | unsigned int ioctl, unsigned long arg) | |
2196 | { | |
2197 | struct kvm *kvm = filp->private_data; | |
2198 | int r; | |
2199 | ||
2200 | if (kvm->mm != current->mm) | |
2201 | return -EIO; | |
2202 | switch (ioctl) { | |
2203 | case KVM_GET_DIRTY_LOG: { | |
2204 | struct compat_kvm_dirty_log compat_log; | |
2205 | struct kvm_dirty_log log; | |
2206 | ||
2207 | r = -EFAULT; | |
2208 | if (copy_from_user(&compat_log, (void __user *)arg, | |
2209 | sizeof(compat_log))) | |
2210 | goto out; | |
2211 | log.slot = compat_log.slot; | |
2212 | log.padding1 = compat_log.padding1; | |
2213 | log.padding2 = compat_log.padding2; | |
2214 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
2215 | ||
2216 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
2217 | if (r) | |
2218 | goto out; | |
2219 | break; | |
2220 | } | |
2221 | default: | |
2222 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
2223 | } | |
2224 | ||
2225 | out: | |
2226 | return r; | |
2227 | } | |
2228 | #endif | |
2229 | ||
e4a533a4 | 2230 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 2231 | { |
777b3f49 MT |
2232 | struct page *page[1]; |
2233 | unsigned long addr; | |
2234 | int npages; | |
2235 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 2236 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 2237 | |
777b3f49 MT |
2238 | addr = gfn_to_hva(kvm, gfn); |
2239 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 2240 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2241 | |
2242 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
2243 | NULL); | |
2244 | if (unlikely(npages != 1)) | |
e4a533a4 | 2245 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2246 | |
2247 | vmf->page = page[0]; | |
e4a533a4 | 2248 | return 0; |
f17abe9a AK |
2249 | } |
2250 | ||
f0f37e2f | 2251 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 2252 | .fault = kvm_vm_fault, |
f17abe9a AK |
2253 | }; |
2254 | ||
2255 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2256 | { | |
2257 | vma->vm_ops = &kvm_vm_vm_ops; | |
2258 | return 0; | |
2259 | } | |
2260 | ||
3d3aab1b | 2261 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
2262 | .release = kvm_vm_release, |
2263 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
2264 | #ifdef CONFIG_COMPAT |
2265 | .compat_ioctl = kvm_vm_compat_ioctl, | |
2266 | #endif | |
f17abe9a | 2267 | .mmap = kvm_vm_mmap, |
6038f373 | 2268 | .llseek = noop_llseek, |
f17abe9a AK |
2269 | }; |
2270 | ||
e08b9637 | 2271 | static int kvm_dev_ioctl_create_vm(unsigned long type) |
f17abe9a | 2272 | { |
aac87636 | 2273 | int r; |
f17abe9a AK |
2274 | struct kvm *kvm; |
2275 | ||
e08b9637 | 2276 | kvm = kvm_create_vm(type); |
d6d28168 AK |
2277 | if (IS_ERR(kvm)) |
2278 | return PTR_ERR(kvm); | |
6ce5a090 TY |
2279 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2280 | r = kvm_coalesced_mmio_init(kvm); | |
2281 | if (r < 0) { | |
2282 | kvm_put_kvm(kvm); | |
2283 | return r; | |
2284 | } | |
2285 | #endif | |
aac87636 HC |
2286 | r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2287 | if (r < 0) | |
66c0b394 | 2288 | kvm_put_kvm(kvm); |
f17abe9a | 2289 | |
aac87636 | 2290 | return r; |
f17abe9a AK |
2291 | } |
2292 | ||
1a811b61 AK |
2293 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2294 | { | |
2295 | switch (arg) { | |
ca9edaee | 2296 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2297 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2298 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
2299 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2300 | case KVM_CAP_SET_BOOT_CPU_ID: | |
2301 | #endif | |
a9c7399d | 2302 | case KVM_CAP_INTERNAL_ERROR_DATA: |
07975ad3 JK |
2303 | #ifdef CONFIG_HAVE_KVM_MSI |
2304 | case KVM_CAP_SIGNAL_MSI: | |
2305 | #endif | |
1a811b61 | 2306 | return 1; |
9900b4b4 | 2307 | #ifdef KVM_CAP_IRQ_ROUTING |
399ec807 | 2308 | case KVM_CAP_IRQ_ROUTING: |
36463146 | 2309 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2310 | #endif |
1a811b61 AK |
2311 | default: |
2312 | break; | |
2313 | } | |
2314 | return kvm_dev_ioctl_check_extension(arg); | |
2315 | } | |
2316 | ||
f17abe9a AK |
2317 | static long kvm_dev_ioctl(struct file *filp, |
2318 | unsigned int ioctl, unsigned long arg) | |
2319 | { | |
07c45a36 | 2320 | long r = -EINVAL; |
f17abe9a AK |
2321 | |
2322 | switch (ioctl) { | |
2323 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2324 | r = -EINVAL; |
2325 | if (arg) | |
2326 | goto out; | |
f17abe9a AK |
2327 | r = KVM_API_VERSION; |
2328 | break; | |
2329 | case KVM_CREATE_VM: | |
e08b9637 | 2330 | r = kvm_dev_ioctl_create_vm(arg); |
f17abe9a | 2331 | break; |
018d00d2 | 2332 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2333 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2334 | break; |
07c45a36 AK |
2335 | case KVM_GET_VCPU_MMAP_SIZE: |
2336 | r = -EINVAL; | |
2337 | if (arg) | |
2338 | goto out; | |
adb1ff46 AK |
2339 | r = PAGE_SIZE; /* struct kvm_run */ |
2340 | #ifdef CONFIG_X86 | |
2341 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2342 | #endif |
2343 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2344 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2345 | #endif |
07c45a36 | 2346 | break; |
d4c9ff2d FEL |
2347 | case KVM_TRACE_ENABLE: |
2348 | case KVM_TRACE_PAUSE: | |
2349 | case KVM_TRACE_DISABLE: | |
2023a29c | 2350 | r = -EOPNOTSUPP; |
d4c9ff2d | 2351 | break; |
6aa8b732 | 2352 | default: |
043405e1 | 2353 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2354 | } |
2355 | out: | |
2356 | return r; | |
2357 | } | |
2358 | ||
6aa8b732 | 2359 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2360 | .unlocked_ioctl = kvm_dev_ioctl, |
2361 | .compat_ioctl = kvm_dev_ioctl, | |
6038f373 | 2362 | .llseek = noop_llseek, |
6aa8b732 AK |
2363 | }; |
2364 | ||
2365 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2366 | KVM_MINOR, |
6aa8b732 AK |
2367 | "kvm", |
2368 | &kvm_chardev_ops, | |
2369 | }; | |
2370 | ||
75b7127c | 2371 | static void hardware_enable_nolock(void *junk) |
1b6c0168 AK |
2372 | { |
2373 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 2374 | int r; |
1b6c0168 | 2375 | |
7f59f492 | 2376 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2377 | return; |
10474ae8 | 2378 | |
7f59f492 | 2379 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
2380 | |
2381 | r = kvm_arch_hardware_enable(NULL); | |
2382 | ||
2383 | if (r) { | |
2384 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
2385 | atomic_inc(&hardware_enable_failed); | |
2386 | printk(KERN_INFO "kvm: enabling virtualization on " | |
2387 | "CPU%d failed\n", cpu); | |
2388 | } | |
1b6c0168 AK |
2389 | } |
2390 | ||
75b7127c TY |
2391 | static void hardware_enable(void *junk) |
2392 | { | |
e935b837 | 2393 | raw_spin_lock(&kvm_lock); |
75b7127c | 2394 | hardware_enable_nolock(junk); |
e935b837 | 2395 | raw_spin_unlock(&kvm_lock); |
75b7127c TY |
2396 | } |
2397 | ||
2398 | static void hardware_disable_nolock(void *junk) | |
1b6c0168 AK |
2399 | { |
2400 | int cpu = raw_smp_processor_id(); | |
2401 | ||
7f59f492 | 2402 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2403 | return; |
7f59f492 | 2404 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2405 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2406 | } |
2407 | ||
75b7127c TY |
2408 | static void hardware_disable(void *junk) |
2409 | { | |
e935b837 | 2410 | raw_spin_lock(&kvm_lock); |
75b7127c | 2411 | hardware_disable_nolock(junk); |
e935b837 | 2412 | raw_spin_unlock(&kvm_lock); |
75b7127c TY |
2413 | } |
2414 | ||
10474ae8 AG |
2415 | static void hardware_disable_all_nolock(void) |
2416 | { | |
2417 | BUG_ON(!kvm_usage_count); | |
2418 | ||
2419 | kvm_usage_count--; | |
2420 | if (!kvm_usage_count) | |
75b7127c | 2421 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
10474ae8 AG |
2422 | } |
2423 | ||
2424 | static void hardware_disable_all(void) | |
2425 | { | |
e935b837 | 2426 | raw_spin_lock(&kvm_lock); |
10474ae8 | 2427 | hardware_disable_all_nolock(); |
e935b837 | 2428 | raw_spin_unlock(&kvm_lock); |
10474ae8 AG |
2429 | } |
2430 | ||
2431 | static int hardware_enable_all(void) | |
2432 | { | |
2433 | int r = 0; | |
2434 | ||
e935b837 | 2435 | raw_spin_lock(&kvm_lock); |
10474ae8 AG |
2436 | |
2437 | kvm_usage_count++; | |
2438 | if (kvm_usage_count == 1) { | |
2439 | atomic_set(&hardware_enable_failed, 0); | |
75b7127c | 2440 | on_each_cpu(hardware_enable_nolock, NULL, 1); |
10474ae8 AG |
2441 | |
2442 | if (atomic_read(&hardware_enable_failed)) { | |
2443 | hardware_disable_all_nolock(); | |
2444 | r = -EBUSY; | |
2445 | } | |
2446 | } | |
2447 | ||
e935b837 | 2448 | raw_spin_unlock(&kvm_lock); |
10474ae8 AG |
2449 | |
2450 | return r; | |
2451 | } | |
2452 | ||
774c47f1 AK |
2453 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2454 | void *v) | |
2455 | { | |
2456 | int cpu = (long)v; | |
2457 | ||
10474ae8 AG |
2458 | if (!kvm_usage_count) |
2459 | return NOTIFY_OK; | |
2460 | ||
1a6f4d7f | 2461 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2462 | switch (val) { |
cec9ad27 | 2463 | case CPU_DYING: |
6ec8a856 AK |
2464 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2465 | cpu); | |
2466 | hardware_disable(NULL); | |
2467 | break; | |
da908f2f | 2468 | case CPU_STARTING: |
43934a38 JK |
2469 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2470 | cpu); | |
da908f2f | 2471 | hardware_enable(NULL); |
774c47f1 AK |
2472 | break; |
2473 | } | |
2474 | return NOTIFY_OK; | |
2475 | } | |
2476 | ||
4ecac3fd | 2477 | |
b7c4145b | 2478 | asmlinkage void kvm_spurious_fault(void) |
4ecac3fd | 2479 | { |
4ecac3fd AK |
2480 | /* Fault while not rebooting. We want the trace. */ |
2481 | BUG(); | |
2482 | } | |
b7c4145b | 2483 | EXPORT_SYMBOL_GPL(kvm_spurious_fault); |
4ecac3fd | 2484 | |
9a2b85c6 | 2485 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2486 | void *v) |
9a2b85c6 | 2487 | { |
8e1c1815 SY |
2488 | /* |
2489 | * Some (well, at least mine) BIOSes hang on reboot if | |
2490 | * in vmx root mode. | |
2491 | * | |
2492 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2493 | */ | |
2494 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2495 | kvm_rebooting = true; | |
75b7127c | 2496 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
9a2b85c6 RR |
2497 | return NOTIFY_OK; |
2498 | } | |
2499 | ||
2500 | static struct notifier_block kvm_reboot_notifier = { | |
2501 | .notifier_call = kvm_reboot, | |
2502 | .priority = 0, | |
2503 | }; | |
2504 | ||
e93f8a0f | 2505 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
2506 | { |
2507 | int i; | |
2508 | ||
2509 | for (i = 0; i < bus->dev_count; i++) { | |
743eeb0b | 2510 | struct kvm_io_device *pos = bus->range[i].dev; |
2eeb2e94 GH |
2511 | |
2512 | kvm_iodevice_destructor(pos); | |
2513 | } | |
e93f8a0f | 2514 | kfree(bus); |
2eeb2e94 GH |
2515 | } |
2516 | ||
743eeb0b SL |
2517 | int kvm_io_bus_sort_cmp(const void *p1, const void *p2) |
2518 | { | |
2519 | const struct kvm_io_range *r1 = p1; | |
2520 | const struct kvm_io_range *r2 = p2; | |
2521 | ||
2522 | if (r1->addr < r2->addr) | |
2523 | return -1; | |
2524 | if (r1->addr + r1->len > r2->addr + r2->len) | |
2525 | return 1; | |
2526 | return 0; | |
2527 | } | |
2528 | ||
2529 | int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev, | |
2530 | gpa_t addr, int len) | |
2531 | { | |
743eeb0b SL |
2532 | bus->range[bus->dev_count++] = (struct kvm_io_range) { |
2533 | .addr = addr, | |
2534 | .len = len, | |
2535 | .dev = dev, | |
2536 | }; | |
2537 | ||
2538 | sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range), | |
2539 | kvm_io_bus_sort_cmp, NULL); | |
2540 | ||
2541 | return 0; | |
2542 | } | |
2543 | ||
2544 | int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus, | |
2545 | gpa_t addr, int len) | |
2546 | { | |
2547 | struct kvm_io_range *range, key; | |
2548 | int off; | |
2549 | ||
2550 | key = (struct kvm_io_range) { | |
2551 | .addr = addr, | |
2552 | .len = len, | |
2553 | }; | |
2554 | ||
2555 | range = bsearch(&key, bus->range, bus->dev_count, | |
2556 | sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp); | |
2557 | if (range == NULL) | |
2558 | return -ENOENT; | |
2559 | ||
2560 | off = range - bus->range; | |
2561 | ||
2562 | while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0) | |
2563 | off--; | |
2564 | ||
2565 | return off; | |
2566 | } | |
2567 | ||
bda9020e | 2568 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2569 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2570 | int len, const void *val) |
2eeb2e94 | 2571 | { |
743eeb0b | 2572 | int idx; |
90d83dc3 | 2573 | struct kvm_io_bus *bus; |
743eeb0b SL |
2574 | struct kvm_io_range range; |
2575 | ||
2576 | range = (struct kvm_io_range) { | |
2577 | .addr = addr, | |
2578 | .len = len, | |
2579 | }; | |
90d83dc3 LJ |
2580 | |
2581 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
743eeb0b SL |
2582 | idx = kvm_io_bus_get_first_dev(bus, addr, len); |
2583 | if (idx < 0) | |
2584 | return -EOPNOTSUPP; | |
2585 | ||
2586 | while (idx < bus->dev_count && | |
2587 | kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) { | |
2588 | if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val)) | |
bda9020e | 2589 | return 0; |
743eeb0b SL |
2590 | idx++; |
2591 | } | |
2592 | ||
bda9020e MT |
2593 | return -EOPNOTSUPP; |
2594 | } | |
2eeb2e94 | 2595 | |
bda9020e | 2596 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2597 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2598 | int len, void *val) | |
bda9020e | 2599 | { |
743eeb0b | 2600 | int idx; |
90d83dc3 | 2601 | struct kvm_io_bus *bus; |
743eeb0b SL |
2602 | struct kvm_io_range range; |
2603 | ||
2604 | range = (struct kvm_io_range) { | |
2605 | .addr = addr, | |
2606 | .len = len, | |
2607 | }; | |
e93f8a0f | 2608 | |
90d83dc3 | 2609 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
743eeb0b SL |
2610 | idx = kvm_io_bus_get_first_dev(bus, addr, len); |
2611 | if (idx < 0) | |
2612 | return -EOPNOTSUPP; | |
2613 | ||
2614 | while (idx < bus->dev_count && | |
2615 | kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) { | |
2616 | if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val)) | |
bda9020e | 2617 | return 0; |
743eeb0b SL |
2618 | idx++; |
2619 | } | |
2620 | ||
bda9020e | 2621 | return -EOPNOTSUPP; |
2eeb2e94 GH |
2622 | } |
2623 | ||
79fac95e | 2624 | /* Caller must hold slots_lock. */ |
743eeb0b SL |
2625 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2626 | int len, struct kvm_io_device *dev) | |
6c474694 | 2627 | { |
e93f8a0f | 2628 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2629 | |
e93f8a0f | 2630 | bus = kvm->buses[bus_idx]; |
a1300716 | 2631 | if (bus->dev_count > NR_IOBUS_DEVS - 1) |
090b7aff | 2632 | return -ENOSPC; |
2eeb2e94 | 2633 | |
a1300716 AK |
2634 | new_bus = kzalloc(sizeof(*bus) + ((bus->dev_count + 1) * |
2635 | sizeof(struct kvm_io_range)), GFP_KERNEL); | |
e93f8a0f MT |
2636 | if (!new_bus) |
2637 | return -ENOMEM; | |
a1300716 AK |
2638 | memcpy(new_bus, bus, sizeof(*bus) + (bus->dev_count * |
2639 | sizeof(struct kvm_io_range))); | |
743eeb0b | 2640 | kvm_io_bus_insert_dev(new_bus, dev, addr, len); |
e93f8a0f MT |
2641 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); |
2642 | synchronize_srcu_expedited(&kvm->srcu); | |
2643 | kfree(bus); | |
090b7aff GH |
2644 | |
2645 | return 0; | |
2646 | } | |
2647 | ||
79fac95e | 2648 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2649 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2650 | struct kvm_io_device *dev) | |
090b7aff | 2651 | { |
e93f8a0f MT |
2652 | int i, r; |
2653 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2654 | |
cdfca7b3 | 2655 | bus = kvm->buses[bus_idx]; |
e93f8a0f | 2656 | r = -ENOENT; |
a1300716 AK |
2657 | for (i = 0; i < bus->dev_count; i++) |
2658 | if (bus->range[i].dev == dev) { | |
e93f8a0f | 2659 | r = 0; |
090b7aff GH |
2660 | break; |
2661 | } | |
e93f8a0f | 2662 | |
a1300716 | 2663 | if (r) |
e93f8a0f | 2664 | return r; |
a1300716 AK |
2665 | |
2666 | new_bus = kzalloc(sizeof(*bus) + ((bus->dev_count - 1) * | |
2667 | sizeof(struct kvm_io_range)), GFP_KERNEL); | |
2668 | if (!new_bus) | |
2669 | return -ENOMEM; | |
2670 | ||
2671 | memcpy(new_bus, bus, sizeof(*bus) + i * sizeof(struct kvm_io_range)); | |
2672 | new_bus->dev_count--; | |
2673 | memcpy(new_bus->range + i, bus->range + i + 1, | |
2674 | (new_bus->dev_count - i) * sizeof(struct kvm_io_range)); | |
e93f8a0f MT |
2675 | |
2676 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2677 | synchronize_srcu_expedited(&kvm->srcu); | |
2678 | kfree(bus); | |
2679 | return r; | |
2eeb2e94 GH |
2680 | } |
2681 | ||
774c47f1 AK |
2682 | static struct notifier_block kvm_cpu_notifier = { |
2683 | .notifier_call = kvm_cpu_hotplug, | |
774c47f1 AK |
2684 | }; |
2685 | ||
8b88b099 | 2686 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2687 | { |
2688 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2689 | struct kvm *kvm; |
2690 | ||
8b88b099 | 2691 | *val = 0; |
e935b837 | 2692 | raw_spin_lock(&kvm_lock); |
ba1389b7 | 2693 | list_for_each_entry(kvm, &vm_list, vm_list) |
8b88b099 | 2694 | *val += *(u32 *)((void *)kvm + offset); |
e935b837 | 2695 | raw_spin_unlock(&kvm_lock); |
8b88b099 | 2696 | return 0; |
ba1389b7 AK |
2697 | } |
2698 | ||
2699 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2700 | ||
8b88b099 | 2701 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2702 | { |
2703 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2704 | struct kvm *kvm; |
2705 | struct kvm_vcpu *vcpu; | |
2706 | int i; | |
2707 | ||
8b88b099 | 2708 | *val = 0; |
e935b837 | 2709 | raw_spin_lock(&kvm_lock); |
1165f5fe | 2710 | list_for_each_entry(kvm, &vm_list, vm_list) |
988a2cae GN |
2711 | kvm_for_each_vcpu(i, vcpu, kvm) |
2712 | *val += *(u32 *)((void *)vcpu + offset); | |
2713 | ||
e935b837 | 2714 | raw_spin_unlock(&kvm_lock); |
8b88b099 | 2715 | return 0; |
1165f5fe AK |
2716 | } |
2717 | ||
ba1389b7 AK |
2718 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2719 | ||
828c0950 | 2720 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2721 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2722 | [KVM_STAT_VM] = &vm_stat_fops, | |
2723 | }; | |
1165f5fe | 2724 | |
4f69b680 | 2725 | static int kvm_init_debug(void) |
6aa8b732 | 2726 | { |
4f69b680 | 2727 | int r = -EFAULT; |
6aa8b732 AK |
2728 | struct kvm_stats_debugfs_item *p; |
2729 | ||
76f7c879 | 2730 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
4f69b680 H |
2731 | if (kvm_debugfs_dir == NULL) |
2732 | goto out; | |
2733 | ||
2734 | for (p = debugfs_entries; p->name; ++p) { | |
76f7c879 | 2735 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2736 | (void *)(long)p->offset, |
ba1389b7 | 2737 | stat_fops[p->kind]); |
4f69b680 H |
2738 | if (p->dentry == NULL) |
2739 | goto out_dir; | |
2740 | } | |
2741 | ||
2742 | return 0; | |
2743 | ||
2744 | out_dir: | |
2745 | debugfs_remove_recursive(kvm_debugfs_dir); | |
2746 | out: | |
2747 | return r; | |
6aa8b732 AK |
2748 | } |
2749 | ||
2750 | static void kvm_exit_debug(void) | |
2751 | { | |
2752 | struct kvm_stats_debugfs_item *p; | |
2753 | ||
2754 | for (p = debugfs_entries; p->name; ++p) | |
2755 | debugfs_remove(p->dentry); | |
76f7c879 | 2756 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2757 | } |
2758 | ||
fb3600cc | 2759 | static int kvm_suspend(void) |
59ae6c6b | 2760 | { |
10474ae8 | 2761 | if (kvm_usage_count) |
75b7127c | 2762 | hardware_disable_nolock(NULL); |
59ae6c6b AK |
2763 | return 0; |
2764 | } | |
2765 | ||
fb3600cc | 2766 | static void kvm_resume(void) |
59ae6c6b | 2767 | { |
ca84d1a2 | 2768 | if (kvm_usage_count) { |
e935b837 | 2769 | WARN_ON(raw_spin_is_locked(&kvm_lock)); |
75b7127c | 2770 | hardware_enable_nolock(NULL); |
ca84d1a2 | 2771 | } |
59ae6c6b AK |
2772 | } |
2773 | ||
fb3600cc | 2774 | static struct syscore_ops kvm_syscore_ops = { |
59ae6c6b AK |
2775 | .suspend = kvm_suspend, |
2776 | .resume = kvm_resume, | |
2777 | }; | |
2778 | ||
15ad7146 AK |
2779 | static inline |
2780 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2781 | { | |
2782 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2783 | } | |
2784 | ||
2785 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2786 | { | |
2787 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2788 | ||
e9b11c17 | 2789 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2790 | } |
2791 | ||
2792 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2793 | struct task_struct *next) | |
2794 | { | |
2795 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2796 | ||
e9b11c17 | 2797 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2798 | } |
2799 | ||
0ee75bea | 2800 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2801 | struct module *module) |
6aa8b732 AK |
2802 | { |
2803 | int r; | |
002c7f7c | 2804 | int cpu; |
6aa8b732 | 2805 | |
f8c16bba ZX |
2806 | r = kvm_arch_init(opaque); |
2807 | if (r) | |
d2308784 | 2808 | goto out_fail; |
cb498ea2 | 2809 | |
8437a617 | 2810 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2811 | r = -ENOMEM; |
2812 | goto out_free_0; | |
2813 | } | |
2814 | ||
e9b11c17 | 2815 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2816 | if (r < 0) |
7f59f492 | 2817 | goto out_free_0a; |
6aa8b732 | 2818 | |
002c7f7c YS |
2819 | for_each_online_cpu(cpu) { |
2820 | smp_call_function_single(cpu, | |
e9b11c17 | 2821 | kvm_arch_check_processor_compat, |
8691e5a8 | 2822 | &r, 1); |
002c7f7c | 2823 | if (r < 0) |
d2308784 | 2824 | goto out_free_1; |
002c7f7c YS |
2825 | } |
2826 | ||
774c47f1 AK |
2827 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2828 | if (r) | |
d2308784 | 2829 | goto out_free_2; |
6aa8b732 AK |
2830 | register_reboot_notifier(&kvm_reboot_notifier); |
2831 | ||
c16f862d | 2832 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2833 | if (!vcpu_align) |
2834 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2835 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2836 | 0, NULL); |
c16f862d RR |
2837 | if (!kvm_vcpu_cache) { |
2838 | r = -ENOMEM; | |
fb3600cc | 2839 | goto out_free_3; |
c16f862d RR |
2840 | } |
2841 | ||
af585b92 GN |
2842 | r = kvm_async_pf_init(); |
2843 | if (r) | |
2844 | goto out_free; | |
2845 | ||
6aa8b732 | 2846 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2847 | kvm_vm_fops.owner = module; |
2848 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2849 | |
2850 | r = misc_register(&kvm_dev); | |
2851 | if (r) { | |
d77c26fc | 2852 | printk(KERN_ERR "kvm: misc device register failed\n"); |
af585b92 | 2853 | goto out_unreg; |
6aa8b732 AK |
2854 | } |
2855 | ||
fb3600cc RW |
2856 | register_syscore_ops(&kvm_syscore_ops); |
2857 | ||
15ad7146 AK |
2858 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2859 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2860 | ||
4f69b680 H |
2861 | r = kvm_init_debug(); |
2862 | if (r) { | |
2863 | printk(KERN_ERR "kvm: create debugfs files failed\n"); | |
2864 | goto out_undebugfs; | |
2865 | } | |
0ea4ed8e | 2866 | |
c7addb90 | 2867 | return 0; |
6aa8b732 | 2868 | |
4f69b680 H |
2869 | out_undebugfs: |
2870 | unregister_syscore_ops(&kvm_syscore_ops); | |
af585b92 GN |
2871 | out_unreg: |
2872 | kvm_async_pf_deinit(); | |
6aa8b732 | 2873 | out_free: |
c16f862d | 2874 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2875 | out_free_3: |
6aa8b732 | 2876 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2877 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2878 | out_free_2: |
d2308784 | 2879 | out_free_1: |
e9b11c17 | 2880 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2881 | out_free_0a: |
2882 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2883 | out_free_0: |
f8c16bba | 2884 | kvm_arch_exit(); |
d2308784 | 2885 | out_fail: |
6aa8b732 AK |
2886 | return r; |
2887 | } | |
cb498ea2 | 2888 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2889 | |
cb498ea2 | 2890 | void kvm_exit(void) |
6aa8b732 | 2891 | { |
0ea4ed8e | 2892 | kvm_exit_debug(); |
6aa8b732 | 2893 | misc_deregister(&kvm_dev); |
c16f862d | 2894 | kmem_cache_destroy(kvm_vcpu_cache); |
af585b92 | 2895 | kvm_async_pf_deinit(); |
fb3600cc | 2896 | unregister_syscore_ops(&kvm_syscore_ops); |
6aa8b732 | 2897 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2898 | unregister_cpu_notifier(&kvm_cpu_notifier); |
75b7127c | 2899 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
e9b11c17 | 2900 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2901 | kvm_arch_exit(); |
7f59f492 | 2902 | free_cpumask_var(cpus_hardware_enabled); |
6aa8b732 | 2903 | } |
cb498ea2 | 2904 | EXPORT_SYMBOL_GPL(kvm_exit); |