]>
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. | |
8 | * | |
9 | * Authors: | |
10 | * Avi Kivity <avi@qumranet.com> | |
11 | * Yaniv Kamay <yaniv@qumranet.com> | |
12 | * | |
13 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
14 | * the COPYING file in the top-level directory. | |
15 | * | |
16 | */ | |
17 | ||
e2174021 | 18 | #include "iodev.h" |
6aa8b732 | 19 | |
edf88417 | 20 | #include <linux/kvm_host.h> |
6aa8b732 AK |
21 | #include <linux/kvm.h> |
22 | #include <linux/module.h> | |
23 | #include <linux/errno.h> | |
6aa8b732 AK |
24 | #include <linux/percpu.h> |
25 | #include <linux/gfp.h> | |
6aa8b732 AK |
26 | #include <linux/mm.h> |
27 | #include <linux/miscdevice.h> | |
28 | #include <linux/vmalloc.h> | |
6aa8b732 | 29 | #include <linux/reboot.h> |
6aa8b732 AK |
30 | #include <linux/debugfs.h> |
31 | #include <linux/highmem.h> | |
32 | #include <linux/file.h> | |
59ae6c6b | 33 | #include <linux/sysdev.h> |
774c47f1 | 34 | #include <linux/cpu.h> |
e8edc6e0 | 35 | #include <linux/sched.h> |
d9e368d6 AK |
36 | #include <linux/cpumask.h> |
37 | #include <linux/smp.h> | |
d6d28168 | 38 | #include <linux/anon_inodes.h> |
04d2cc77 | 39 | #include <linux/profile.h> |
7aa81cc0 | 40 | #include <linux/kvm_para.h> |
6fc138d2 | 41 | #include <linux/pagemap.h> |
8d4e1288 | 42 | #include <linux/mman.h> |
35149e21 | 43 | #include <linux/swap.h> |
e56d532f | 44 | #include <linux/bitops.h> |
547de29e | 45 | #include <linux/spinlock.h> |
6ff5894c | 46 | #include <linux/compat.h> |
6aa8b732 | 47 | |
e495606d | 48 | #include <asm/processor.h> |
e495606d AK |
49 | #include <asm/io.h> |
50 | #include <asm/uaccess.h> | |
3e021bf5 | 51 | #include <asm/pgtable.h> |
c8240bd6 | 52 | #include <asm-generic/bitops/le.h> |
6aa8b732 | 53 | |
5f94c174 LV |
54 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
55 | #include "coalesced_mmio.h" | |
56 | #endif | |
57 | ||
229456fc MT |
58 | #define CREATE_TRACE_POINTS |
59 | #include <trace/events/kvm.h> | |
60 | ||
6aa8b732 AK |
61 | MODULE_AUTHOR("Qumranet"); |
62 | MODULE_LICENSE("GPL"); | |
63 | ||
fa40a821 MT |
64 | /* |
65 | * Ordering of locks: | |
66 | * | |
22fc0294 | 67 | * kvm->slots_lock --> kvm->lock --> kvm->irq_lock |
fa40a821 MT |
68 | */ |
69 | ||
e9b11c17 ZX |
70 | DEFINE_SPINLOCK(kvm_lock); |
71 | LIST_HEAD(vm_list); | |
133de902 | 72 | |
7f59f492 | 73 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
74 | static int kvm_usage_count = 0; |
75 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 76 | |
c16f862d RR |
77 | struct kmem_cache *kvm_vcpu_cache; |
78 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 79 | |
15ad7146 AK |
80 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
81 | ||
76f7c879 | 82 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 83 | |
bccf2150 AK |
84 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
85 | unsigned long arg); | |
10474ae8 AG |
86 | static int hardware_enable_all(void); |
87 | static void hardware_disable_all(void); | |
bccf2150 | 88 | |
e8ba5d31 | 89 | static bool kvm_rebooting; |
4ecac3fd | 90 | |
54dee993 MT |
91 | static bool largepages_enabled = true; |
92 | ||
c77fb9dc | 93 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 94 | { |
fc5659c8 JR |
95 | if (pfn_valid(pfn)) { |
96 | struct page *page = compound_head(pfn_to_page(pfn)); | |
97 | return PageReserved(page); | |
98 | } | |
cbff90a7 BAY |
99 | |
100 | return true; | |
101 | } | |
102 | ||
bccf2150 AK |
103 | /* |
104 | * Switches to specified vcpu, until a matching vcpu_put() | |
105 | */ | |
313a3dc7 | 106 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 107 | { |
15ad7146 AK |
108 | int cpu; |
109 | ||
bccf2150 | 110 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
111 | cpu = get_cpu(); |
112 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 113 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 114 | put_cpu(); |
6aa8b732 AK |
115 | } |
116 | ||
313a3dc7 | 117 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 118 | { |
15ad7146 | 119 | preempt_disable(); |
313a3dc7 | 120 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
121 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
122 | preempt_enable(); | |
6aa8b732 AK |
123 | mutex_unlock(&vcpu->mutex); |
124 | } | |
125 | ||
d9e368d6 AK |
126 | static void ack_flush(void *_completed) |
127 | { | |
d9e368d6 AK |
128 | } |
129 | ||
49846896 | 130 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 131 | { |
597a5f55 | 132 | int i, cpu, me; |
6ef7a1bc RR |
133 | cpumask_var_t cpus; |
134 | bool called = true; | |
d9e368d6 | 135 | struct kvm_vcpu *vcpu; |
d9e368d6 | 136 | |
79f55997 | 137 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 138 | |
84261923 | 139 | spin_lock(&kvm->requests_lock); |
e601e3be | 140 | me = smp_processor_id(); |
988a2cae | 141 | kvm_for_each_vcpu(i, vcpu, kvm) { |
49846896 | 142 | if (test_and_set_bit(req, &vcpu->requests)) |
d9e368d6 AK |
143 | continue; |
144 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
145 | if (cpus != NULL && cpu != -1 && cpu != me) |
146 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 147 | } |
6ef7a1bc RR |
148 | if (unlikely(cpus == NULL)) |
149 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
150 | else if (!cpumask_empty(cpus)) | |
151 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
152 | else | |
153 | called = false; | |
84261923 | 154 | spin_unlock(&kvm->requests_lock); |
6ef7a1bc | 155 | free_cpumask_var(cpus); |
49846896 | 156 | return called; |
d9e368d6 AK |
157 | } |
158 | ||
49846896 | 159 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 160 | { |
49846896 RR |
161 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
162 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
163 | } |
164 | ||
49846896 RR |
165 | void kvm_reload_remote_mmus(struct kvm *kvm) |
166 | { | |
167 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
168 | } | |
2e53d63a | 169 | |
fb3f0f51 RR |
170 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
171 | { | |
172 | struct page *page; | |
173 | int r; | |
174 | ||
175 | mutex_init(&vcpu->mutex); | |
176 | vcpu->cpu = -1; | |
fb3f0f51 RR |
177 | vcpu->kvm = kvm; |
178 | vcpu->vcpu_id = id; | |
b6958ce4 | 179 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
180 | |
181 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
182 | if (!page) { | |
183 | r = -ENOMEM; | |
184 | goto fail; | |
185 | } | |
186 | vcpu->run = page_address(page); | |
187 | ||
e9b11c17 | 188 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 189 | if (r < 0) |
e9b11c17 | 190 | goto fail_free_run; |
fb3f0f51 RR |
191 | return 0; |
192 | ||
fb3f0f51 RR |
193 | fail_free_run: |
194 | free_page((unsigned long)vcpu->run); | |
195 | fail: | |
76fafa5e | 196 | return r; |
fb3f0f51 RR |
197 | } |
198 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
199 | ||
200 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
201 | { | |
e9b11c17 | 202 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
203 | free_page((unsigned long)vcpu->run); |
204 | } | |
205 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
206 | ||
e930bffe AA |
207 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
208 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
209 | { | |
210 | return container_of(mn, struct kvm, mmu_notifier); | |
211 | } | |
212 | ||
213 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
214 | struct mm_struct *mm, | |
215 | unsigned long address) | |
216 | { | |
217 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
218 | int need_tlb_flush; | |
219 | ||
220 | /* | |
221 | * When ->invalidate_page runs, the linux pte has been zapped | |
222 | * already but the page is still allocated until | |
223 | * ->invalidate_page returns. So if we increase the sequence | |
224 | * here the kvm page fault will notice if the spte can't be | |
225 | * established because the page is going to be freed. If | |
226 | * instead the kvm page fault establishes the spte before | |
227 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
228 | * before returning. | |
229 | * | |
230 | * The sequence increase only need to be seen at spin_unlock | |
231 | * time, and not at spin_lock time. | |
232 | * | |
233 | * Increasing the sequence after the spin_unlock would be | |
234 | * unsafe because the kvm page fault could then establish the | |
235 | * pte after kvm_unmap_hva returned, without noticing the page | |
236 | * is going to be freed. | |
237 | */ | |
238 | spin_lock(&kvm->mmu_lock); | |
239 | kvm->mmu_notifier_seq++; | |
240 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
241 | spin_unlock(&kvm->mmu_lock); | |
242 | ||
243 | /* we've to flush the tlb before the pages can be freed */ | |
244 | if (need_tlb_flush) | |
245 | kvm_flush_remote_tlbs(kvm); | |
246 | ||
247 | } | |
248 | ||
3da0dd43 IE |
249 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
250 | struct mm_struct *mm, | |
251 | unsigned long address, | |
252 | pte_t pte) | |
253 | { | |
254 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
255 | ||
256 | spin_lock(&kvm->mmu_lock); | |
257 | kvm->mmu_notifier_seq++; | |
258 | kvm_set_spte_hva(kvm, address, pte); | |
259 | spin_unlock(&kvm->mmu_lock); | |
260 | } | |
261 | ||
e930bffe AA |
262 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
263 | struct mm_struct *mm, | |
264 | unsigned long start, | |
265 | unsigned long end) | |
266 | { | |
267 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
268 | int need_tlb_flush = 0; | |
269 | ||
270 | spin_lock(&kvm->mmu_lock); | |
271 | /* | |
272 | * The count increase must become visible at unlock time as no | |
273 | * spte can be established without taking the mmu_lock and | |
274 | * count is also read inside the mmu_lock critical section. | |
275 | */ | |
276 | kvm->mmu_notifier_count++; | |
277 | for (; start < end; start += PAGE_SIZE) | |
278 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
279 | spin_unlock(&kvm->mmu_lock); | |
280 | ||
281 | /* we've to flush the tlb before the pages can be freed */ | |
282 | if (need_tlb_flush) | |
283 | kvm_flush_remote_tlbs(kvm); | |
284 | } | |
285 | ||
286 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
287 | struct mm_struct *mm, | |
288 | unsigned long start, | |
289 | unsigned long end) | |
290 | { | |
291 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
292 | ||
293 | spin_lock(&kvm->mmu_lock); | |
294 | /* | |
295 | * This sequence increase will notify the kvm page fault that | |
296 | * the page that is going to be mapped in the spte could have | |
297 | * been freed. | |
298 | */ | |
299 | kvm->mmu_notifier_seq++; | |
300 | /* | |
301 | * The above sequence increase must be visible before the | |
302 | * below count decrease but both values are read by the kvm | |
303 | * page fault under mmu_lock spinlock so we don't need to add | |
304 | * a smb_wmb() here in between the two. | |
305 | */ | |
306 | kvm->mmu_notifier_count--; | |
307 | spin_unlock(&kvm->mmu_lock); | |
308 | ||
309 | BUG_ON(kvm->mmu_notifier_count < 0); | |
310 | } | |
311 | ||
312 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
313 | struct mm_struct *mm, | |
314 | unsigned long address) | |
315 | { | |
316 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
317 | int young; | |
318 | ||
319 | spin_lock(&kvm->mmu_lock); | |
320 | young = kvm_age_hva(kvm, address); | |
321 | spin_unlock(&kvm->mmu_lock); | |
322 | ||
323 | if (young) | |
324 | kvm_flush_remote_tlbs(kvm); | |
325 | ||
326 | return young; | |
327 | } | |
328 | ||
85db06e5 MT |
329 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
330 | struct mm_struct *mm) | |
331 | { | |
332 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
333 | kvm_arch_flush_shadow(kvm); | |
334 | } | |
335 | ||
e930bffe AA |
336 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
337 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
338 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
339 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
340 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
3da0dd43 | 341 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 342 | .release = kvm_mmu_notifier_release, |
e930bffe AA |
343 | }; |
344 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ | |
345 | ||
f17abe9a | 346 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 347 | { |
10474ae8 | 348 | int r = 0; |
d19a9cd2 | 349 | struct kvm *kvm = kvm_arch_create_vm(); |
5f94c174 LV |
350 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
351 | struct page *page; | |
352 | #endif | |
6aa8b732 | 353 | |
d19a9cd2 ZX |
354 | if (IS_ERR(kvm)) |
355 | goto out; | |
10474ae8 AG |
356 | |
357 | r = hardware_enable_all(); | |
358 | if (r) | |
359 | goto out_err_nodisable; | |
360 | ||
75858a84 AK |
361 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
362 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 363 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 364 | #endif |
6aa8b732 | 365 | |
5f94c174 LV |
366 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
367 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
368 | if (!page) { | |
10474ae8 AG |
369 | r = -ENOMEM; |
370 | goto out_err; | |
5f94c174 LV |
371 | } |
372 | kvm->coalesced_mmio_ring = | |
373 | (struct kvm_coalesced_mmio_ring *)page_address(page); | |
374 | #endif | |
375 | ||
e930bffe AA |
376 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
377 | { | |
e930bffe | 378 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; |
10474ae8 AG |
379 | r = mmu_notifier_register(&kvm->mmu_notifier, current->mm); |
380 | if (r) { | |
e930bffe AA |
381 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
382 | put_page(page); | |
383 | #endif | |
10474ae8 | 384 | goto out_err; |
e930bffe AA |
385 | } |
386 | } | |
387 | #endif | |
388 | ||
6d4e4c4f AK |
389 | kvm->mm = current->mm; |
390 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 391 | spin_lock_init(&kvm->mmu_lock); |
84261923 | 392 | spin_lock_init(&kvm->requests_lock); |
74906345 | 393 | kvm_io_bus_init(&kvm->pio_bus); |
d34e6b17 | 394 | kvm_eventfd_init(kvm); |
11ec2804 | 395 | mutex_init(&kvm->lock); |
60eead79 | 396 | mutex_init(&kvm->irq_lock); |
2eeb2e94 | 397 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 398 | init_rwsem(&kvm->slots_lock); |
d39f13b0 | 399 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
400 | spin_lock(&kvm_lock); |
401 | list_add(&kvm->vm_list, &vm_list); | |
402 | spin_unlock(&kvm_lock); | |
5f94c174 LV |
403 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
404 | kvm_coalesced_mmio_init(kvm); | |
405 | #endif | |
d19a9cd2 | 406 | out: |
f17abe9a | 407 | return kvm; |
10474ae8 AG |
408 | |
409 | out_err: | |
410 | hardware_disable_all(); | |
411 | out_err_nodisable: | |
412 | kfree(kvm); | |
413 | return ERR_PTR(r); | |
f17abe9a AK |
414 | } |
415 | ||
6aa8b732 AK |
416 | /* |
417 | * Free any memory in @free but not in @dont. | |
418 | */ | |
419 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
420 | struct kvm_memory_slot *dont) | |
421 | { | |
ec04b260 JR |
422 | int i; |
423 | ||
290fc38d IE |
424 | if (!dont || free->rmap != dont->rmap) |
425 | vfree(free->rmap); | |
6aa8b732 AK |
426 | |
427 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
428 | vfree(free->dirty_bitmap); | |
429 | ||
ec04b260 JR |
430 | |
431 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
432 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
433 | vfree(free->lpage_info[i]); | |
434 | free->lpage_info[i] = NULL; | |
435 | } | |
436 | } | |
05da4558 | 437 | |
6aa8b732 | 438 | free->npages = 0; |
8b6d44c7 | 439 | free->dirty_bitmap = NULL; |
8d4e1288 | 440 | free->rmap = NULL; |
6aa8b732 AK |
441 | } |
442 | ||
d19a9cd2 | 443 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
444 | { |
445 | int i; | |
446 | ||
447 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 448 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
449 | } |
450 | ||
f17abe9a AK |
451 | static void kvm_destroy_vm(struct kvm *kvm) |
452 | { | |
6d4e4c4f AK |
453 | struct mm_struct *mm = kvm->mm; |
454 | ||
ad8ba2cd | 455 | kvm_arch_sync_events(kvm); |
133de902 AK |
456 | spin_lock(&kvm_lock); |
457 | list_del(&kvm->vm_list); | |
458 | spin_unlock(&kvm_lock); | |
399ec807 | 459 | kvm_free_irq_routing(kvm); |
74906345 | 460 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 461 | kvm_io_bus_destroy(&kvm->mmio_bus); |
5f94c174 LV |
462 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
463 | if (kvm->coalesced_mmio_ring != NULL) | |
464 | free_page((unsigned long)kvm->coalesced_mmio_ring); | |
e930bffe AA |
465 | #endif |
466 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) | |
467 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
468 | #else |
469 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 470 | #endif |
d19a9cd2 | 471 | kvm_arch_destroy_vm(kvm); |
10474ae8 | 472 | hardware_disable_all(); |
6d4e4c4f | 473 | mmdrop(mm); |
f17abe9a AK |
474 | } |
475 | ||
d39f13b0 IE |
476 | void kvm_get_kvm(struct kvm *kvm) |
477 | { | |
478 | atomic_inc(&kvm->users_count); | |
479 | } | |
480 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
481 | ||
482 | void kvm_put_kvm(struct kvm *kvm) | |
483 | { | |
484 | if (atomic_dec_and_test(&kvm->users_count)) | |
485 | kvm_destroy_vm(kvm); | |
486 | } | |
487 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
488 | ||
489 | ||
f17abe9a AK |
490 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
491 | { | |
492 | struct kvm *kvm = filp->private_data; | |
493 | ||
721eecbf GH |
494 | kvm_irqfd_release(kvm); |
495 | ||
d39f13b0 | 496 | kvm_put_kvm(kvm); |
6aa8b732 AK |
497 | return 0; |
498 | } | |
499 | ||
6aa8b732 AK |
500 | /* |
501 | * Allocate some memory and give it an address in the guest physical address | |
502 | * space. | |
503 | * | |
504 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 505 | * |
10589a46 | 506 | * Must be called holding mmap_sem for write. |
6aa8b732 | 507 | */ |
f78e0e2e SY |
508 | int __kvm_set_memory_region(struct kvm *kvm, |
509 | struct kvm_userspace_memory_region *mem, | |
510 | int user_alloc) | |
6aa8b732 AK |
511 | { |
512 | int r; | |
513 | gfn_t base_gfn; | |
28bcb112 HC |
514 | unsigned long npages; |
515 | unsigned long i; | |
6aa8b732 AK |
516 | struct kvm_memory_slot *memslot; |
517 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
518 | |
519 | r = -EINVAL; | |
520 | /* General sanity checks */ | |
521 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
522 | goto out; | |
523 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
524 | goto out; | |
e7cacd40 | 525 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 526 | goto out; |
e0d62c7f | 527 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
528 | goto out; |
529 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
530 | goto out; | |
531 | ||
532 | memslot = &kvm->memslots[mem->slot]; | |
533 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
534 | npages = mem->memory_size >> PAGE_SHIFT; | |
535 | ||
536 | if (!npages) | |
537 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
538 | ||
6aa8b732 AK |
539 | new = old = *memslot; |
540 | ||
541 | new.base_gfn = base_gfn; | |
542 | new.npages = npages; | |
543 | new.flags = mem->flags; | |
544 | ||
545 | /* Disallow changing a memory slot's size. */ | |
546 | r = -EINVAL; | |
547 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 548 | goto out_free; |
6aa8b732 AK |
549 | |
550 | /* Check for overlaps */ | |
551 | r = -EEXIST; | |
552 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
553 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
554 | ||
4cd481f6 | 555 | if (s == memslot || !s->npages) |
6aa8b732 AK |
556 | continue; |
557 | if (!((base_gfn + npages <= s->base_gfn) || | |
558 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 559 | goto out_free; |
6aa8b732 | 560 | } |
6aa8b732 | 561 | |
6aa8b732 AK |
562 | /* Free page dirty bitmap if unneeded */ |
563 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 564 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
565 | |
566 | r = -ENOMEM; | |
567 | ||
568 | /* Allocate if a slot is being created */ | |
eff0114a | 569 | #ifndef CONFIG_S390 |
8d4e1288 | 570 | if (npages && !new.rmap) { |
d77c26fc | 571 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
572 | |
573 | if (!new.rmap) | |
f78e0e2e | 574 | goto out_free; |
290fc38d | 575 | |
290fc38d | 576 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 577 | |
80b14b5b | 578 | new.user_alloc = user_alloc; |
604b38ac AA |
579 | /* |
580 | * hva_to_rmmap() serialzies with the mmu_lock and to be | |
581 | * safe it has to ignore memslots with !user_alloc && | |
582 | * !userspace_addr. | |
583 | */ | |
584 | if (user_alloc) | |
585 | new.userspace_addr = mem->userspace_addr; | |
586 | else | |
587 | new.userspace_addr = 0; | |
6aa8b732 | 588 | } |
ec04b260 JR |
589 | if (!npages) |
590 | goto skip_lpage; | |
05da4558 | 591 | |
ec04b260 | 592 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
593 | unsigned long ugfn; |
594 | unsigned long j; | |
595 | int lpages; | |
ec04b260 | 596 | int level = i + 2; |
05da4558 | 597 | |
ec04b260 JR |
598 | /* Avoid unused variable warning if no large pages */ |
599 | (void)level; | |
600 | ||
601 | if (new.lpage_info[i]) | |
602 | continue; | |
603 | ||
604 | lpages = 1 + (base_gfn + npages - 1) / | |
605 | KVM_PAGES_PER_HPAGE(level); | |
606 | lpages -= base_gfn / KVM_PAGES_PER_HPAGE(level); | |
607 | ||
608 | new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i])); | |
609 | ||
610 | if (!new.lpage_info[i]) | |
05da4558 MT |
611 | goto out_free; |
612 | ||
ec04b260 JR |
613 | memset(new.lpage_info[i], 0, |
614 | lpages * sizeof(*new.lpage_info[i])); | |
05da4558 | 615 | |
ec04b260 JR |
616 | if (base_gfn % KVM_PAGES_PER_HPAGE(level)) |
617 | new.lpage_info[i][0].write_count = 1; | |
618 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE(level)) | |
619 | new.lpage_info[i][lpages - 1].write_count = 1; | |
ac04527f AK |
620 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
621 | /* | |
622 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
623 | * other, or if explicitly asked to, disable large page |
624 | * support for this slot | |
ac04527f | 625 | */ |
ec04b260 | 626 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 627 | !largepages_enabled) |
ec04b260 JR |
628 | for (j = 0; j < lpages; ++j) |
629 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 630 | } |
6aa8b732 | 631 | |
ec04b260 JR |
632 | skip_lpage: |
633 | ||
6aa8b732 AK |
634 | /* Allocate page dirty bitmap if needed */ |
635 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
636 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
637 | ||
638 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
639 | if (!new.dirty_bitmap) | |
f78e0e2e | 640 | goto out_free; |
6aa8b732 | 641 | memset(new.dirty_bitmap, 0, dirty_bytes); |
e244584f IE |
642 | if (old.npages) |
643 | kvm_arch_flush_shadow(kvm); | |
6aa8b732 | 644 | } |
3eea8437 CB |
645 | #else /* not defined CONFIG_S390 */ |
646 | new.user_alloc = user_alloc; | |
647 | if (user_alloc) | |
648 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 649 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 650 | |
34d4cb8f MT |
651 | if (!npages) |
652 | kvm_arch_flush_shadow(kvm); | |
653 | ||
604b38ac AA |
654 | spin_lock(&kvm->mmu_lock); |
655 | if (mem->slot >= kvm->nmemslots) | |
656 | kvm->nmemslots = mem->slot + 1; | |
657 | ||
3ad82a7e | 658 | *memslot = new; |
604b38ac | 659 | spin_unlock(&kvm->mmu_lock); |
3ad82a7e | 660 | |
0de10343 ZX |
661 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
662 | if (r) { | |
604b38ac | 663 | spin_lock(&kvm->mmu_lock); |
0de10343 | 664 | *memslot = old; |
604b38ac | 665 | spin_unlock(&kvm->mmu_lock); |
0de10343 | 666 | goto out_free; |
82ce2c96 IE |
667 | } |
668 | ||
6f897248 GC |
669 | kvm_free_physmem_slot(&old, npages ? &new : NULL); |
670 | /* Slot deletion case: we have to update the current slot */ | |
b43b1901 | 671 | spin_lock(&kvm->mmu_lock); |
6f897248 GC |
672 | if (!npages) |
673 | *memslot = old; | |
b43b1901 | 674 | spin_unlock(&kvm->mmu_lock); |
8a98f664 | 675 | #ifdef CONFIG_DMAR |
62c476c7 BAY |
676 | /* map the pages in iommu page table */ |
677 | r = kvm_iommu_map_pages(kvm, base_gfn, npages); | |
678 | if (r) | |
679 | goto out; | |
8a98f664 | 680 | #endif |
6aa8b732 AK |
681 | return 0; |
682 | ||
f78e0e2e | 683 | out_free: |
6aa8b732 AK |
684 | kvm_free_physmem_slot(&new, &old); |
685 | out: | |
686 | return r; | |
210c7c4d IE |
687 | |
688 | } | |
f78e0e2e SY |
689 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
690 | ||
691 | int kvm_set_memory_region(struct kvm *kvm, | |
692 | struct kvm_userspace_memory_region *mem, | |
693 | int user_alloc) | |
694 | { | |
695 | int r; | |
696 | ||
72dc67a6 | 697 | down_write(&kvm->slots_lock); |
f78e0e2e | 698 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 699 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
700 | return r; |
701 | } | |
210c7c4d IE |
702 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
703 | ||
1fe779f8 CO |
704 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
705 | struct | |
706 | kvm_userspace_memory_region *mem, | |
707 | int user_alloc) | |
210c7c4d | 708 | { |
e0d62c7f IE |
709 | if (mem->slot >= KVM_MEMORY_SLOTS) |
710 | return -EINVAL; | |
210c7c4d | 711 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
712 | } |
713 | ||
5bb064dc ZX |
714 | int kvm_get_dirty_log(struct kvm *kvm, |
715 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
716 | { |
717 | struct kvm_memory_slot *memslot; | |
718 | int r, i; | |
719 | int n; | |
720 | unsigned long any = 0; | |
721 | ||
6aa8b732 AK |
722 | r = -EINVAL; |
723 | if (log->slot >= KVM_MEMORY_SLOTS) | |
724 | goto out; | |
725 | ||
726 | memslot = &kvm->memslots[log->slot]; | |
727 | r = -ENOENT; | |
728 | if (!memslot->dirty_bitmap) | |
729 | goto out; | |
730 | ||
cd1a4a98 | 731 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 732 | |
cd1a4a98 | 733 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
734 | any = memslot->dirty_bitmap[i]; |
735 | ||
736 | r = -EFAULT; | |
737 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
738 | goto out; | |
739 | ||
5bb064dc ZX |
740 | if (any) |
741 | *is_dirty = 1; | |
6aa8b732 AK |
742 | |
743 | r = 0; | |
6aa8b732 | 744 | out: |
6aa8b732 AK |
745 | return r; |
746 | } | |
747 | ||
54dee993 MT |
748 | void kvm_disable_largepages(void) |
749 | { | |
750 | largepages_enabled = false; | |
751 | } | |
752 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
753 | ||
cea7bb21 IE |
754 | int is_error_page(struct page *page) |
755 | { | |
756 | return page == bad_page; | |
757 | } | |
758 | EXPORT_SYMBOL_GPL(is_error_page); | |
759 | ||
35149e21 AL |
760 | int is_error_pfn(pfn_t pfn) |
761 | { | |
762 | return pfn == bad_pfn; | |
763 | } | |
764 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
765 | ||
f9d46eb0 IE |
766 | static inline unsigned long bad_hva(void) |
767 | { | |
768 | return PAGE_OFFSET; | |
769 | } | |
770 | ||
771 | int kvm_is_error_hva(unsigned long addr) | |
772 | { | |
773 | return addr == bad_hva(); | |
774 | } | |
775 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
776 | ||
2843099f | 777 | struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
778 | { |
779 | int i; | |
780 | ||
781 | for (i = 0; i < kvm->nmemslots; ++i) { | |
782 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
783 | ||
784 | if (gfn >= memslot->base_gfn | |
785 | && gfn < memslot->base_gfn + memslot->npages) | |
786 | return memslot; | |
787 | } | |
8b6d44c7 | 788 | return NULL; |
6aa8b732 | 789 | } |
2843099f | 790 | EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased); |
e8207547 AK |
791 | |
792 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
793 | { | |
794 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 795 | return gfn_to_memslot_unaliased(kvm, gfn); |
e8207547 | 796 | } |
6aa8b732 | 797 | |
e0d62c7f IE |
798 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
799 | { | |
800 | int i; | |
801 | ||
802 | gfn = unalias_gfn(kvm, gfn); | |
803 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
804 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
805 | ||
806 | if (gfn >= memslot->base_gfn | |
807 | && gfn < memslot->base_gfn + memslot->npages) | |
808 | return 1; | |
809 | } | |
810 | return 0; | |
811 | } | |
812 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
813 | ||
05da4558 | 814 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
815 | { |
816 | struct kvm_memory_slot *slot; | |
817 | ||
818 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 819 | slot = gfn_to_memslot_unaliased(kvm, gfn); |
539cb660 IE |
820 | if (!slot) |
821 | return bad_hva(); | |
822 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
823 | } | |
0d150298 | 824 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 825 | |
35149e21 | 826 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 827 | { |
8d4e1288 | 828 | struct page *page[1]; |
539cb660 | 829 | unsigned long addr; |
8d4e1288 | 830 | int npages; |
2e2e3738 | 831 | pfn_t pfn; |
954bbbc2 | 832 | |
60395224 AK |
833 | might_sleep(); |
834 | ||
539cb660 IE |
835 | addr = gfn_to_hva(kvm, gfn); |
836 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 837 | get_page(bad_page); |
35149e21 | 838 | return page_to_pfn(bad_page); |
8a7ae055 | 839 | } |
8d4e1288 | 840 | |
4c2155ce | 841 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 842 | |
2e2e3738 AL |
843 | if (unlikely(npages != 1)) { |
844 | struct vm_area_struct *vma; | |
845 | ||
4c2155ce | 846 | down_read(¤t->mm->mmap_sem); |
2e2e3738 | 847 | vma = find_vma(current->mm, addr); |
4c2155ce | 848 | |
2e2e3738 AL |
849 | if (vma == NULL || addr < vma->vm_start || |
850 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 851 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
852 | get_page(bad_page); |
853 | return page_to_pfn(bad_page); | |
854 | } | |
855 | ||
856 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 857 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 858 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
859 | } else |
860 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 861 | |
2e2e3738 | 862 | return pfn; |
35149e21 AL |
863 | } |
864 | ||
865 | EXPORT_SYMBOL_GPL(gfn_to_pfn); | |
866 | ||
867 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
868 | { | |
2e2e3738 AL |
869 | pfn_t pfn; |
870 | ||
871 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 872 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
873 | return pfn_to_page(pfn); |
874 | ||
c77fb9dc | 875 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
876 | |
877 | get_page(bad_page); | |
878 | return bad_page; | |
954bbbc2 | 879 | } |
aab61cc0 | 880 | |
954bbbc2 AK |
881 | EXPORT_SYMBOL_GPL(gfn_to_page); |
882 | ||
b4231d61 IE |
883 | void kvm_release_page_clean(struct page *page) |
884 | { | |
35149e21 | 885 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
886 | } |
887 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
888 | ||
35149e21 AL |
889 | void kvm_release_pfn_clean(pfn_t pfn) |
890 | { | |
c77fb9dc | 891 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 892 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
893 | } |
894 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
895 | ||
b4231d61 | 896 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 897 | { |
35149e21 AL |
898 | kvm_release_pfn_dirty(page_to_pfn(page)); |
899 | } | |
900 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
901 | ||
902 | void kvm_release_pfn_dirty(pfn_t pfn) | |
903 | { | |
904 | kvm_set_pfn_dirty(pfn); | |
905 | kvm_release_pfn_clean(pfn); | |
906 | } | |
907 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
908 | ||
909 | void kvm_set_page_dirty(struct page *page) | |
910 | { | |
911 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
912 | } | |
913 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
914 | ||
915 | void kvm_set_pfn_dirty(pfn_t pfn) | |
916 | { | |
c77fb9dc | 917 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
918 | struct page *page = pfn_to_page(pfn); |
919 | if (!PageReserved(page)) | |
920 | SetPageDirty(page); | |
921 | } | |
8a7ae055 | 922 | } |
35149e21 AL |
923 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
924 | ||
925 | void kvm_set_pfn_accessed(pfn_t pfn) | |
926 | { | |
c77fb9dc | 927 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 928 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
929 | } |
930 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
931 | ||
932 | void kvm_get_pfn(pfn_t pfn) | |
933 | { | |
c77fb9dc | 934 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 935 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
936 | } |
937 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 938 | |
195aefde IE |
939 | static int next_segment(unsigned long len, int offset) |
940 | { | |
941 | if (len > PAGE_SIZE - offset) | |
942 | return PAGE_SIZE - offset; | |
943 | else | |
944 | return len; | |
945 | } | |
946 | ||
947 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
948 | int len) | |
949 | { | |
e0506bcb IE |
950 | int r; |
951 | unsigned long addr; | |
195aefde | 952 | |
e0506bcb IE |
953 | addr = gfn_to_hva(kvm, gfn); |
954 | if (kvm_is_error_hva(addr)) | |
955 | return -EFAULT; | |
956 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
957 | if (r) | |
195aefde | 958 | return -EFAULT; |
195aefde IE |
959 | return 0; |
960 | } | |
961 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
962 | ||
963 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
964 | { | |
965 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
966 | int seg; | |
967 | int offset = offset_in_page(gpa); | |
968 | int ret; | |
969 | ||
970 | while ((seg = next_segment(len, offset)) != 0) { | |
971 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
972 | if (ret < 0) | |
973 | return ret; | |
974 | offset = 0; | |
975 | len -= seg; | |
976 | data += seg; | |
977 | ++gfn; | |
978 | } | |
979 | return 0; | |
980 | } | |
981 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
982 | ||
7ec54588 MT |
983 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
984 | unsigned long len) | |
985 | { | |
986 | int r; | |
987 | unsigned long addr; | |
988 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
989 | int offset = offset_in_page(gpa); | |
990 | ||
991 | addr = gfn_to_hva(kvm, gfn); | |
992 | if (kvm_is_error_hva(addr)) | |
993 | return -EFAULT; | |
0aac03f0 | 994 | pagefault_disable(); |
7ec54588 | 995 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 996 | pagefault_enable(); |
7ec54588 MT |
997 | if (r) |
998 | return -EFAULT; | |
999 | return 0; | |
1000 | } | |
1001 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1002 | ||
195aefde IE |
1003 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1004 | int offset, int len) | |
1005 | { | |
e0506bcb IE |
1006 | int r; |
1007 | unsigned long addr; | |
195aefde | 1008 | |
e0506bcb IE |
1009 | addr = gfn_to_hva(kvm, gfn); |
1010 | if (kvm_is_error_hva(addr)) | |
1011 | return -EFAULT; | |
1012 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1013 | if (r) | |
195aefde | 1014 | return -EFAULT; |
195aefde IE |
1015 | mark_page_dirty(kvm, gfn); |
1016 | return 0; | |
1017 | } | |
1018 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1019 | ||
1020 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1021 | unsigned long len) | |
1022 | { | |
1023 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1024 | int seg; | |
1025 | int offset = offset_in_page(gpa); | |
1026 | int ret; | |
1027 | ||
1028 | while ((seg = next_segment(len, offset)) != 0) { | |
1029 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1030 | if (ret < 0) | |
1031 | return ret; | |
1032 | offset = 0; | |
1033 | len -= seg; | |
1034 | data += seg; | |
1035 | ++gfn; | |
1036 | } | |
1037 | return 0; | |
1038 | } | |
1039 | ||
1040 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1041 | { | |
3e021bf5 | 1042 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1043 | } |
1044 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1045 | ||
1046 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1047 | { | |
1048 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1049 | int seg; | |
1050 | int offset = offset_in_page(gpa); | |
1051 | int ret; | |
1052 | ||
1053 | while ((seg = next_segment(len, offset)) != 0) { | |
1054 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1055 | if (ret < 0) | |
1056 | return ret; | |
1057 | offset = 0; | |
1058 | len -= seg; | |
1059 | ++gfn; | |
1060 | } | |
1061 | return 0; | |
1062 | } | |
1063 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1064 | ||
6aa8b732 AK |
1065 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1066 | { | |
31389947 | 1067 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1068 | |
3b6fff19 | 1069 | gfn = unalias_gfn(kvm, gfn); |
2843099f | 1070 | memslot = gfn_to_memslot_unaliased(kvm, gfn); |
7e9d619d RR |
1071 | if (memslot && memslot->dirty_bitmap) { |
1072 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1073 | |
7e9d619d | 1074 | /* avoid RMW */ |
c8240bd6 AG |
1075 | if (!generic_test_le_bit(rel_gfn, memslot->dirty_bitmap)) |
1076 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1077 | } |
1078 | } | |
1079 | ||
b6958ce4 ED |
1080 | /* |
1081 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1082 | */ | |
8776e519 | 1083 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1084 | { |
e5c239cf MT |
1085 | DEFINE_WAIT(wait); |
1086 | ||
1087 | for (;;) { | |
1088 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1089 | ||
a1b37100 | 1090 | if (kvm_arch_vcpu_runnable(vcpu)) { |
d7690175 | 1091 | set_bit(KVM_REQ_UNHALT, &vcpu->requests); |
e5c239cf | 1092 | break; |
d7690175 | 1093 | } |
09cec754 GN |
1094 | if (kvm_cpu_has_pending_timer(vcpu)) |
1095 | break; | |
e5c239cf MT |
1096 | if (signal_pending(current)) |
1097 | break; | |
1098 | ||
b6958ce4 | 1099 | schedule(); |
b6958ce4 | 1100 | } |
d3bef15f | 1101 | |
e5c239cf | 1102 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1103 | } |
1104 | ||
6aa8b732 AK |
1105 | void kvm_resched(struct kvm_vcpu *vcpu) |
1106 | { | |
3fca0365 YD |
1107 | if (!need_resched()) |
1108 | return; | |
6aa8b732 | 1109 | cond_resched(); |
6aa8b732 AK |
1110 | } |
1111 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1112 | ||
d255f4f2 ZE |
1113 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu) |
1114 | { | |
1115 | ktime_t expires; | |
1116 | DEFINE_WAIT(wait); | |
1117 | ||
1118 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1119 | ||
1120 | /* Sleep for 100 us, and hope lock-holder got scheduled */ | |
1121 | expires = ktime_add_ns(ktime_get(), 100000UL); | |
1122 | schedule_hrtimeout(&expires, HRTIMER_MODE_ABS); | |
1123 | ||
1124 | finish_wait(&vcpu->wq, &wait); | |
1125 | } | |
1126 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1127 | ||
e4a533a4 | 1128 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1129 | { |
1130 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1131 | struct page *page; |
1132 | ||
e4a533a4 | 1133 | if (vmf->pgoff == 0) |
039576c0 | 1134 | page = virt_to_page(vcpu->run); |
09566765 | 1135 | #ifdef CONFIG_X86 |
e4a533a4 | 1136 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1137 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1138 | #endif |
1139 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1140 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1141 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1142 | #endif |
039576c0 | 1143 | else |
e4a533a4 | 1144 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1145 | get_page(page); |
e4a533a4 | 1146 | vmf->page = page; |
1147 | return 0; | |
9a2bb7f4 AK |
1148 | } |
1149 | ||
f0f37e2f | 1150 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1151 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1152 | }; |
1153 | ||
1154 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1155 | { | |
1156 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1157 | return 0; | |
1158 | } | |
1159 | ||
bccf2150 AK |
1160 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1161 | { | |
1162 | struct kvm_vcpu *vcpu = filp->private_data; | |
1163 | ||
66c0b394 | 1164 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1165 | return 0; |
1166 | } | |
1167 | ||
3d3aab1b | 1168 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1169 | .release = kvm_vcpu_release, |
1170 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1171 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1172 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1173 | }; |
1174 | ||
1175 | /* | |
1176 | * Allocates an inode for the vcpu. | |
1177 | */ | |
1178 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1179 | { | |
73880c80 | 1180 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, 0); |
bccf2150 AK |
1181 | } |
1182 | ||
c5ea7660 AK |
1183 | /* |
1184 | * Creates some virtual cpus. Good luck creating more than one. | |
1185 | */ | |
73880c80 | 1186 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1187 | { |
1188 | int r; | |
988a2cae | 1189 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1190 | |
73880c80 | 1191 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1192 | if (IS_ERR(vcpu)) |
1193 | return PTR_ERR(vcpu); | |
c5ea7660 | 1194 | |
15ad7146 AK |
1195 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1196 | ||
26e5215f AK |
1197 | r = kvm_arch_vcpu_setup(vcpu); |
1198 | if (r) | |
7d8fece6 | 1199 | return r; |
26e5215f | 1200 | |
11ec2804 | 1201 | mutex_lock(&kvm->lock); |
73880c80 GN |
1202 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1203 | r = -EINVAL; | |
e9b11c17 | 1204 | goto vcpu_destroy; |
fb3f0f51 | 1205 | } |
73880c80 | 1206 | |
988a2cae GN |
1207 | kvm_for_each_vcpu(r, v, kvm) |
1208 | if (v->vcpu_id == id) { | |
73880c80 GN |
1209 | r = -EEXIST; |
1210 | goto vcpu_destroy; | |
1211 | } | |
1212 | ||
1213 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1214 | |
fb3f0f51 | 1215 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1216 | kvm_get_kvm(kvm); |
bccf2150 | 1217 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1218 | if (r < 0) { |
1219 | kvm_put_kvm(kvm); | |
1220 | goto vcpu_destroy; | |
1221 | } | |
1222 | ||
1223 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1224 | smp_wmb(); | |
1225 | atomic_inc(&kvm->online_vcpus); | |
1226 | ||
1227 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1228 | if (kvm->bsp_vcpu_id == id) | |
1229 | kvm->bsp_vcpu = vcpu; | |
1230 | #endif | |
1231 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1232 | return r; |
39c3b86e | 1233 | |
e9b11c17 | 1234 | vcpu_destroy: |
7d8fece6 | 1235 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1236 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1237 | return r; |
1238 | } | |
1239 | ||
1961d276 AK |
1240 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1241 | { | |
1242 | if (sigset) { | |
1243 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1244 | vcpu->sigset_active = 1; | |
1245 | vcpu->sigset = *sigset; | |
1246 | } else | |
1247 | vcpu->sigset_active = 0; | |
1248 | return 0; | |
1249 | } | |
1250 | ||
bccf2150 AK |
1251 | static long kvm_vcpu_ioctl(struct file *filp, |
1252 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1253 | { |
bccf2150 | 1254 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1255 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1256 | int r; |
fa3795a7 DH |
1257 | struct kvm_fpu *fpu = NULL; |
1258 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1259 | |
6d4e4c4f AK |
1260 | if (vcpu->kvm->mm != current->mm) |
1261 | return -EIO; | |
6aa8b732 | 1262 | switch (ioctl) { |
9a2bb7f4 | 1263 | case KVM_RUN: |
f0fe5108 AK |
1264 | r = -EINVAL; |
1265 | if (arg) | |
1266 | goto out; | |
b6c7a5dc | 1267 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1268 | break; |
6aa8b732 | 1269 | case KVM_GET_REGS: { |
3e4bb3ac | 1270 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1271 | |
3e4bb3ac XZ |
1272 | r = -ENOMEM; |
1273 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1274 | if (!kvm_regs) | |
6aa8b732 | 1275 | goto out; |
3e4bb3ac XZ |
1276 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1277 | if (r) | |
1278 | goto out_free1; | |
6aa8b732 | 1279 | r = -EFAULT; |
3e4bb3ac XZ |
1280 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1281 | goto out_free1; | |
6aa8b732 | 1282 | r = 0; |
3e4bb3ac XZ |
1283 | out_free1: |
1284 | kfree(kvm_regs); | |
6aa8b732 AK |
1285 | break; |
1286 | } | |
1287 | case KVM_SET_REGS: { | |
3e4bb3ac | 1288 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1289 | |
3e4bb3ac XZ |
1290 | r = -ENOMEM; |
1291 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1292 | if (!kvm_regs) | |
6aa8b732 | 1293 | goto out; |
3e4bb3ac XZ |
1294 | r = -EFAULT; |
1295 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1296 | goto out_free2; | |
1297 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1298 | if (r) |
3e4bb3ac | 1299 | goto out_free2; |
6aa8b732 | 1300 | r = 0; |
3e4bb3ac XZ |
1301 | out_free2: |
1302 | kfree(kvm_regs); | |
6aa8b732 AK |
1303 | break; |
1304 | } | |
1305 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1306 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1307 | r = -ENOMEM; | |
1308 | if (!kvm_sregs) | |
1309 | goto out; | |
1310 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1311 | if (r) |
1312 | goto out; | |
1313 | r = -EFAULT; | |
fa3795a7 | 1314 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1315 | goto out; |
1316 | r = 0; | |
1317 | break; | |
1318 | } | |
1319 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1320 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1321 | r = -ENOMEM; | |
1322 | if (!kvm_sregs) | |
1323 | goto out; | |
6aa8b732 | 1324 | r = -EFAULT; |
fa3795a7 | 1325 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1326 | goto out; |
fa3795a7 | 1327 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1328 | if (r) |
1329 | goto out; | |
1330 | r = 0; | |
1331 | break; | |
1332 | } | |
62d9f0db MT |
1333 | case KVM_GET_MP_STATE: { |
1334 | struct kvm_mp_state mp_state; | |
1335 | ||
1336 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1337 | if (r) | |
1338 | goto out; | |
1339 | r = -EFAULT; | |
1340 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1341 | goto out; | |
1342 | r = 0; | |
1343 | break; | |
1344 | } | |
1345 | case KVM_SET_MP_STATE: { | |
1346 | struct kvm_mp_state mp_state; | |
1347 | ||
1348 | r = -EFAULT; | |
1349 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1350 | goto out; | |
1351 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1352 | if (r) | |
1353 | goto out; | |
1354 | r = 0; | |
1355 | break; | |
1356 | } | |
6aa8b732 AK |
1357 | case KVM_TRANSLATE: { |
1358 | struct kvm_translation tr; | |
1359 | ||
1360 | r = -EFAULT; | |
2f366987 | 1361 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1362 | goto out; |
8b006791 | 1363 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1364 | if (r) |
1365 | goto out; | |
1366 | r = -EFAULT; | |
2f366987 | 1367 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1368 | goto out; |
1369 | r = 0; | |
1370 | break; | |
1371 | } | |
d0bfb940 JK |
1372 | case KVM_SET_GUEST_DEBUG: { |
1373 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1374 | |
1375 | r = -EFAULT; | |
2f366987 | 1376 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1377 | goto out; |
d0bfb940 | 1378 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1379 | if (r) |
1380 | goto out; | |
1381 | r = 0; | |
1382 | break; | |
1383 | } | |
1961d276 AK |
1384 | case KVM_SET_SIGNAL_MASK: { |
1385 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1386 | struct kvm_signal_mask kvm_sigmask; | |
1387 | sigset_t sigset, *p; | |
1388 | ||
1389 | p = NULL; | |
1390 | if (argp) { | |
1391 | r = -EFAULT; | |
1392 | if (copy_from_user(&kvm_sigmask, argp, | |
1393 | sizeof kvm_sigmask)) | |
1394 | goto out; | |
1395 | r = -EINVAL; | |
1396 | if (kvm_sigmask.len != sizeof sigset) | |
1397 | goto out; | |
1398 | r = -EFAULT; | |
1399 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1400 | sizeof sigset)) | |
1401 | goto out; | |
1402 | p = &sigset; | |
1403 | } | |
1404 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1405 | break; | |
1406 | } | |
b8836737 | 1407 | case KVM_GET_FPU: { |
fa3795a7 DH |
1408 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1409 | r = -ENOMEM; | |
1410 | if (!fpu) | |
1411 | goto out; | |
1412 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1413 | if (r) |
1414 | goto out; | |
1415 | r = -EFAULT; | |
fa3795a7 | 1416 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1417 | goto out; |
1418 | r = 0; | |
1419 | break; | |
1420 | } | |
1421 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1422 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1423 | r = -ENOMEM; | |
1424 | if (!fpu) | |
1425 | goto out; | |
b8836737 | 1426 | r = -EFAULT; |
fa3795a7 | 1427 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1428 | goto out; |
fa3795a7 | 1429 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1430 | if (r) |
1431 | goto out; | |
1432 | r = 0; | |
1433 | break; | |
1434 | } | |
bccf2150 | 1435 | default: |
313a3dc7 | 1436 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1437 | } |
1438 | out: | |
fa3795a7 DH |
1439 | kfree(fpu); |
1440 | kfree(kvm_sregs); | |
bccf2150 AK |
1441 | return r; |
1442 | } | |
1443 | ||
1444 | static long kvm_vm_ioctl(struct file *filp, | |
1445 | unsigned int ioctl, unsigned long arg) | |
1446 | { | |
1447 | struct kvm *kvm = filp->private_data; | |
1448 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1449 | int r; |
bccf2150 | 1450 | |
6d4e4c4f AK |
1451 | if (kvm->mm != current->mm) |
1452 | return -EIO; | |
bccf2150 AK |
1453 | switch (ioctl) { |
1454 | case KVM_CREATE_VCPU: | |
1455 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1456 | if (r < 0) | |
1457 | goto out; | |
1458 | break; | |
6fc138d2 IE |
1459 | case KVM_SET_USER_MEMORY_REGION: { |
1460 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1461 | ||
1462 | r = -EFAULT; | |
1463 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1464 | sizeof kvm_userspace_mem)) | |
1465 | goto out; | |
1466 | ||
1467 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1468 | if (r) |
1469 | goto out; | |
1470 | break; | |
1471 | } | |
1472 | case KVM_GET_DIRTY_LOG: { | |
1473 | struct kvm_dirty_log log; | |
1474 | ||
1475 | r = -EFAULT; | |
2f366987 | 1476 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1477 | goto out; |
2c6f5df9 | 1478 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1479 | if (r) |
1480 | goto out; | |
1481 | break; | |
1482 | } | |
5f94c174 LV |
1483 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1484 | case KVM_REGISTER_COALESCED_MMIO: { | |
1485 | struct kvm_coalesced_mmio_zone zone; | |
1486 | r = -EFAULT; | |
1487 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1488 | goto out; | |
1489 | r = -ENXIO; | |
1490 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); | |
1491 | if (r) | |
1492 | goto out; | |
1493 | r = 0; | |
1494 | break; | |
1495 | } | |
1496 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1497 | struct kvm_coalesced_mmio_zone zone; | |
1498 | r = -EFAULT; | |
1499 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1500 | goto out; | |
1501 | r = -ENXIO; | |
1502 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); | |
1503 | if (r) | |
1504 | goto out; | |
1505 | r = 0; | |
1506 | break; | |
1507 | } | |
1508 | #endif | |
721eecbf GH |
1509 | case KVM_IRQFD: { |
1510 | struct kvm_irqfd data; | |
1511 | ||
1512 | r = -EFAULT; | |
1513 | if (copy_from_user(&data, argp, sizeof data)) | |
1514 | goto out; | |
1515 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1516 | break; | |
1517 | } | |
d34e6b17 GH |
1518 | case KVM_IOEVENTFD: { |
1519 | struct kvm_ioeventfd data; | |
1520 | ||
1521 | r = -EFAULT; | |
1522 | if (copy_from_user(&data, argp, sizeof data)) | |
1523 | goto out; | |
1524 | r = kvm_ioeventfd(kvm, &data); | |
1525 | break; | |
1526 | } | |
73880c80 GN |
1527 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1528 | case KVM_SET_BOOT_CPU_ID: | |
1529 | r = 0; | |
894a9c55 | 1530 | mutex_lock(&kvm->lock); |
73880c80 GN |
1531 | if (atomic_read(&kvm->online_vcpus) != 0) |
1532 | r = -EBUSY; | |
1533 | else | |
1534 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1535 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1536 | break; |
1537 | #endif | |
f17abe9a | 1538 | default: |
1fe779f8 | 1539 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1540 | if (r == -ENOTTY) |
1541 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1542 | } |
1543 | out: | |
1544 | return r; | |
1545 | } | |
1546 | ||
6ff5894c AB |
1547 | #ifdef CONFIG_COMPAT |
1548 | struct compat_kvm_dirty_log { | |
1549 | __u32 slot; | |
1550 | __u32 padding1; | |
1551 | union { | |
1552 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1553 | __u64 padding2; | |
1554 | }; | |
1555 | }; | |
1556 | ||
1557 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1558 | unsigned int ioctl, unsigned long arg) | |
1559 | { | |
1560 | struct kvm *kvm = filp->private_data; | |
1561 | int r; | |
1562 | ||
1563 | if (kvm->mm != current->mm) | |
1564 | return -EIO; | |
1565 | switch (ioctl) { | |
1566 | case KVM_GET_DIRTY_LOG: { | |
1567 | struct compat_kvm_dirty_log compat_log; | |
1568 | struct kvm_dirty_log log; | |
1569 | ||
1570 | r = -EFAULT; | |
1571 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1572 | sizeof(compat_log))) | |
1573 | goto out; | |
1574 | log.slot = compat_log.slot; | |
1575 | log.padding1 = compat_log.padding1; | |
1576 | log.padding2 = compat_log.padding2; | |
1577 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
1578 | ||
1579 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
1580 | if (r) | |
1581 | goto out; | |
1582 | break; | |
1583 | } | |
1584 | default: | |
1585 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
1586 | } | |
1587 | ||
1588 | out: | |
1589 | return r; | |
1590 | } | |
1591 | #endif | |
1592 | ||
e4a533a4 | 1593 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1594 | { |
777b3f49 MT |
1595 | struct page *page[1]; |
1596 | unsigned long addr; | |
1597 | int npages; | |
1598 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1599 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1600 | |
777b3f49 MT |
1601 | addr = gfn_to_hva(kvm, gfn); |
1602 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1603 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1604 | |
1605 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1606 | NULL); | |
1607 | if (unlikely(npages != 1)) | |
e4a533a4 | 1608 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1609 | |
1610 | vmf->page = page[0]; | |
e4a533a4 | 1611 | return 0; |
f17abe9a AK |
1612 | } |
1613 | ||
f0f37e2f | 1614 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 1615 | .fault = kvm_vm_fault, |
f17abe9a AK |
1616 | }; |
1617 | ||
1618 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1619 | { | |
1620 | vma->vm_ops = &kvm_vm_vm_ops; | |
1621 | return 0; | |
1622 | } | |
1623 | ||
3d3aab1b | 1624 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1625 | .release = kvm_vm_release, |
1626 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
1627 | #ifdef CONFIG_COMPAT |
1628 | .compat_ioctl = kvm_vm_compat_ioctl, | |
1629 | #endif | |
f17abe9a AK |
1630 | .mmap = kvm_vm_mmap, |
1631 | }; | |
1632 | ||
1633 | static int kvm_dev_ioctl_create_vm(void) | |
1634 | { | |
2030a42c | 1635 | int fd; |
f17abe9a AK |
1636 | struct kvm *kvm; |
1637 | ||
f17abe9a | 1638 | kvm = kvm_create_vm(); |
d6d28168 AK |
1639 | if (IS_ERR(kvm)) |
1640 | return PTR_ERR(kvm); | |
7d9dbca3 | 1641 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, 0); |
2030a42c | 1642 | if (fd < 0) |
66c0b394 | 1643 | kvm_put_kvm(kvm); |
f17abe9a | 1644 | |
f17abe9a | 1645 | return fd; |
f17abe9a AK |
1646 | } |
1647 | ||
1a811b61 AK |
1648 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
1649 | { | |
1650 | switch (arg) { | |
ca9edaee | 1651 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 1652 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 1653 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
1654 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1655 | case KVM_CAP_SET_BOOT_CPU_ID: | |
1656 | #endif | |
a9c7399d | 1657 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 1658 | return 1; |
399ec807 AK |
1659 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
1660 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 1661 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 1662 | #endif |
1a811b61 AK |
1663 | default: |
1664 | break; | |
1665 | } | |
1666 | return kvm_dev_ioctl_check_extension(arg); | |
1667 | } | |
1668 | ||
f17abe9a AK |
1669 | static long kvm_dev_ioctl(struct file *filp, |
1670 | unsigned int ioctl, unsigned long arg) | |
1671 | { | |
07c45a36 | 1672 | long r = -EINVAL; |
f17abe9a AK |
1673 | |
1674 | switch (ioctl) { | |
1675 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1676 | r = -EINVAL; |
1677 | if (arg) | |
1678 | goto out; | |
f17abe9a AK |
1679 | r = KVM_API_VERSION; |
1680 | break; | |
1681 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1682 | r = -EINVAL; |
1683 | if (arg) | |
1684 | goto out; | |
f17abe9a AK |
1685 | r = kvm_dev_ioctl_create_vm(); |
1686 | break; | |
018d00d2 | 1687 | case KVM_CHECK_EXTENSION: |
1a811b61 | 1688 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 1689 | break; |
07c45a36 AK |
1690 | case KVM_GET_VCPU_MMAP_SIZE: |
1691 | r = -EINVAL; | |
1692 | if (arg) | |
1693 | goto out; | |
adb1ff46 AK |
1694 | r = PAGE_SIZE; /* struct kvm_run */ |
1695 | #ifdef CONFIG_X86 | |
1696 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
1697 | #endif |
1698 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1699 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 1700 | #endif |
07c45a36 | 1701 | break; |
d4c9ff2d FEL |
1702 | case KVM_TRACE_ENABLE: |
1703 | case KVM_TRACE_PAUSE: | |
1704 | case KVM_TRACE_DISABLE: | |
2023a29c | 1705 | r = -EOPNOTSUPP; |
d4c9ff2d | 1706 | break; |
6aa8b732 | 1707 | default: |
043405e1 | 1708 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1709 | } |
1710 | out: | |
1711 | return r; | |
1712 | } | |
1713 | ||
6aa8b732 | 1714 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1715 | .unlocked_ioctl = kvm_dev_ioctl, |
1716 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1717 | }; |
1718 | ||
1719 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1720 | KVM_MINOR, |
6aa8b732 AK |
1721 | "kvm", |
1722 | &kvm_chardev_ops, | |
1723 | }; | |
1724 | ||
1b6c0168 AK |
1725 | static void hardware_enable(void *junk) |
1726 | { | |
1727 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 1728 | int r; |
1b6c0168 | 1729 | |
7f59f492 | 1730 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1731 | return; |
10474ae8 | 1732 | |
7f59f492 | 1733 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
1734 | |
1735 | r = kvm_arch_hardware_enable(NULL); | |
1736 | ||
1737 | if (r) { | |
1738 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
1739 | atomic_inc(&hardware_enable_failed); | |
1740 | printk(KERN_INFO "kvm: enabling virtualization on " | |
1741 | "CPU%d failed\n", cpu); | |
1742 | } | |
1b6c0168 AK |
1743 | } |
1744 | ||
1745 | static void hardware_disable(void *junk) | |
1746 | { | |
1747 | int cpu = raw_smp_processor_id(); | |
1748 | ||
7f59f492 | 1749 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1750 | return; |
7f59f492 | 1751 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 1752 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1753 | } |
1754 | ||
10474ae8 AG |
1755 | static void hardware_disable_all_nolock(void) |
1756 | { | |
1757 | BUG_ON(!kvm_usage_count); | |
1758 | ||
1759 | kvm_usage_count--; | |
1760 | if (!kvm_usage_count) | |
1761 | on_each_cpu(hardware_disable, NULL, 1); | |
1762 | } | |
1763 | ||
1764 | static void hardware_disable_all(void) | |
1765 | { | |
1766 | spin_lock(&kvm_lock); | |
1767 | hardware_disable_all_nolock(); | |
1768 | spin_unlock(&kvm_lock); | |
1769 | } | |
1770 | ||
1771 | static int hardware_enable_all(void) | |
1772 | { | |
1773 | int r = 0; | |
1774 | ||
1775 | spin_lock(&kvm_lock); | |
1776 | ||
1777 | kvm_usage_count++; | |
1778 | if (kvm_usage_count == 1) { | |
1779 | atomic_set(&hardware_enable_failed, 0); | |
1780 | on_each_cpu(hardware_enable, NULL, 1); | |
1781 | ||
1782 | if (atomic_read(&hardware_enable_failed)) { | |
1783 | hardware_disable_all_nolock(); | |
1784 | r = -EBUSY; | |
1785 | } | |
1786 | } | |
1787 | ||
1788 | spin_unlock(&kvm_lock); | |
1789 | ||
1790 | return r; | |
1791 | } | |
1792 | ||
774c47f1 AK |
1793 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1794 | void *v) | |
1795 | { | |
1796 | int cpu = (long)v; | |
1797 | ||
10474ae8 AG |
1798 | if (!kvm_usage_count) |
1799 | return NOTIFY_OK; | |
1800 | ||
1a6f4d7f | 1801 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1802 | switch (val) { |
cec9ad27 | 1803 | case CPU_DYING: |
6ec8a856 AK |
1804 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1805 | cpu); | |
1806 | hardware_disable(NULL); | |
1807 | break; | |
774c47f1 | 1808 | case CPU_UP_CANCELED: |
43934a38 JK |
1809 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1810 | cpu); | |
8691e5a8 | 1811 | smp_call_function_single(cpu, hardware_disable, NULL, 1); |
774c47f1 | 1812 | break; |
43934a38 JK |
1813 | case CPU_ONLINE: |
1814 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1815 | cpu); | |
8691e5a8 | 1816 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
1817 | break; |
1818 | } | |
1819 | return NOTIFY_OK; | |
1820 | } | |
1821 | ||
4ecac3fd AK |
1822 | |
1823 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
1824 | { | |
1825 | if (kvm_rebooting) | |
1826 | /* spin while reset goes on */ | |
1827 | while (true) | |
1828 | ; | |
1829 | /* Fault while not rebooting. We want the trace. */ | |
1830 | BUG(); | |
1831 | } | |
1832 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
1833 | ||
9a2b85c6 | 1834 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1835 | void *v) |
9a2b85c6 | 1836 | { |
8e1c1815 SY |
1837 | /* |
1838 | * Some (well, at least mine) BIOSes hang on reboot if | |
1839 | * in vmx root mode. | |
1840 | * | |
1841 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
1842 | */ | |
1843 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1844 | kvm_rebooting = true; | |
1845 | on_each_cpu(hardware_disable, NULL, 1); | |
9a2b85c6 RR |
1846 | return NOTIFY_OK; |
1847 | } | |
1848 | ||
1849 | static struct notifier_block kvm_reboot_notifier = { | |
1850 | .notifier_call = kvm_reboot, | |
1851 | .priority = 0, | |
1852 | }; | |
1853 | ||
2eeb2e94 GH |
1854 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
1855 | { | |
1856 | memset(bus, 0, sizeof(*bus)); | |
1857 | } | |
1858 | ||
1859 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
1860 | { | |
1861 | int i; | |
1862 | ||
1863 | for (i = 0; i < bus->dev_count; i++) { | |
1864 | struct kvm_io_device *pos = bus->devs[i]; | |
1865 | ||
1866 | kvm_iodevice_destructor(pos); | |
1867 | } | |
1868 | } | |
1869 | ||
bda9020e MT |
1870 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
1871 | int kvm_io_bus_write(struct kvm_io_bus *bus, gpa_t addr, | |
1872 | int len, const void *val) | |
2eeb2e94 GH |
1873 | { |
1874 | int i; | |
bda9020e MT |
1875 | for (i = 0; i < bus->dev_count; i++) |
1876 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
1877 | return 0; | |
1878 | return -EOPNOTSUPP; | |
1879 | } | |
2eeb2e94 | 1880 | |
bda9020e MT |
1881 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
1882 | int kvm_io_bus_read(struct kvm_io_bus *bus, gpa_t addr, int len, void *val) | |
1883 | { | |
1884 | int i; | |
1885 | for (i = 0; i < bus->dev_count; i++) | |
1886 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
1887 | return 0; | |
1888 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
1889 | } |
1890 | ||
090b7aff | 1891 | int kvm_io_bus_register_dev(struct kvm *kvm, struct kvm_io_bus *bus, |
6c474694 MT |
1892 | struct kvm_io_device *dev) |
1893 | { | |
090b7aff GH |
1894 | int ret; |
1895 | ||
6c474694 | 1896 | down_write(&kvm->slots_lock); |
090b7aff | 1897 | ret = __kvm_io_bus_register_dev(bus, dev); |
6c474694 | 1898 | up_write(&kvm->slots_lock); |
090b7aff GH |
1899 | |
1900 | return ret; | |
6c474694 MT |
1901 | } |
1902 | ||
1903 | /* An unlocked version. Caller must have write lock on slots_lock. */ | |
090b7aff GH |
1904 | int __kvm_io_bus_register_dev(struct kvm_io_bus *bus, |
1905 | struct kvm_io_device *dev) | |
2eeb2e94 | 1906 | { |
090b7aff GH |
1907 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
1908 | return -ENOSPC; | |
2eeb2e94 GH |
1909 | |
1910 | bus->devs[bus->dev_count++] = dev; | |
090b7aff GH |
1911 | |
1912 | return 0; | |
1913 | } | |
1914 | ||
1915 | void kvm_io_bus_unregister_dev(struct kvm *kvm, | |
1916 | struct kvm_io_bus *bus, | |
1917 | struct kvm_io_device *dev) | |
1918 | { | |
1919 | down_write(&kvm->slots_lock); | |
1920 | __kvm_io_bus_unregister_dev(bus, dev); | |
1921 | up_write(&kvm->slots_lock); | |
1922 | } | |
1923 | ||
1924 | /* An unlocked version. Caller must have write lock on slots_lock. */ | |
1925 | void __kvm_io_bus_unregister_dev(struct kvm_io_bus *bus, | |
1926 | struct kvm_io_device *dev) | |
1927 | { | |
1928 | int i; | |
1929 | ||
1930 | for (i = 0; i < bus->dev_count; i++) | |
1931 | if (bus->devs[i] == dev) { | |
1932 | bus->devs[i] = bus->devs[--bus->dev_count]; | |
1933 | break; | |
1934 | } | |
2eeb2e94 GH |
1935 | } |
1936 | ||
774c47f1 AK |
1937 | static struct notifier_block kvm_cpu_notifier = { |
1938 | .notifier_call = kvm_cpu_hotplug, | |
1939 | .priority = 20, /* must be > scheduler priority */ | |
1940 | }; | |
1941 | ||
8b88b099 | 1942 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
1943 | { |
1944 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
1945 | struct kvm *kvm; |
1946 | ||
8b88b099 | 1947 | *val = 0; |
ba1389b7 AK |
1948 | spin_lock(&kvm_lock); |
1949 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 1950 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 1951 | spin_unlock(&kvm_lock); |
8b88b099 | 1952 | return 0; |
ba1389b7 AK |
1953 | } |
1954 | ||
1955 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
1956 | ||
8b88b099 | 1957 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
1958 | { |
1959 | unsigned offset = (long)_offset; | |
1165f5fe AK |
1960 | struct kvm *kvm; |
1961 | struct kvm_vcpu *vcpu; | |
1962 | int i; | |
1963 | ||
8b88b099 | 1964 | *val = 0; |
1165f5fe AK |
1965 | spin_lock(&kvm_lock); |
1966 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
1967 | kvm_for_each_vcpu(i, vcpu, kvm) |
1968 | *val += *(u32 *)((void *)vcpu + offset); | |
1969 | ||
1165f5fe | 1970 | spin_unlock(&kvm_lock); |
8b88b099 | 1971 | return 0; |
1165f5fe AK |
1972 | } |
1973 | ||
ba1389b7 AK |
1974 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
1975 | ||
828c0950 | 1976 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
1977 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
1978 | [KVM_STAT_VM] = &vm_stat_fops, | |
1979 | }; | |
1165f5fe | 1980 | |
a16b043c | 1981 | static void kvm_init_debug(void) |
6aa8b732 AK |
1982 | { |
1983 | struct kvm_stats_debugfs_item *p; | |
1984 | ||
76f7c879 | 1985 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 1986 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 1987 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 1988 | (void *)(long)p->offset, |
ba1389b7 | 1989 | stat_fops[p->kind]); |
6aa8b732 AK |
1990 | } |
1991 | ||
1992 | static void kvm_exit_debug(void) | |
1993 | { | |
1994 | struct kvm_stats_debugfs_item *p; | |
1995 | ||
1996 | for (p = debugfs_entries; p->name; ++p) | |
1997 | debugfs_remove(p->dentry); | |
76f7c879 | 1998 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
1999 | } |
2000 | ||
59ae6c6b AK |
2001 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2002 | { | |
10474ae8 AG |
2003 | if (kvm_usage_count) |
2004 | hardware_disable(NULL); | |
59ae6c6b AK |
2005 | return 0; |
2006 | } | |
2007 | ||
2008 | static int kvm_resume(struct sys_device *dev) | |
2009 | { | |
10474ae8 AG |
2010 | if (kvm_usage_count) |
2011 | hardware_enable(NULL); | |
59ae6c6b AK |
2012 | return 0; |
2013 | } | |
2014 | ||
2015 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2016 | .name = "kvm", |
59ae6c6b AK |
2017 | .suspend = kvm_suspend, |
2018 | .resume = kvm_resume, | |
2019 | }; | |
2020 | ||
2021 | static struct sys_device kvm_sysdev = { | |
2022 | .id = 0, | |
2023 | .cls = &kvm_sysdev_class, | |
2024 | }; | |
2025 | ||
cea7bb21 | 2026 | struct page *bad_page; |
35149e21 | 2027 | pfn_t bad_pfn; |
6aa8b732 | 2028 | |
15ad7146 AK |
2029 | static inline |
2030 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2031 | { | |
2032 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2033 | } | |
2034 | ||
2035 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2036 | { | |
2037 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2038 | ||
e9b11c17 | 2039 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2040 | } |
2041 | ||
2042 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2043 | struct task_struct *next) | |
2044 | { | |
2045 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2046 | ||
e9b11c17 | 2047 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2048 | } |
2049 | ||
f8c16bba | 2050 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 2051 | struct module *module) |
6aa8b732 AK |
2052 | { |
2053 | int r; | |
002c7f7c | 2054 | int cpu; |
6aa8b732 | 2055 | |
f8c16bba ZX |
2056 | r = kvm_arch_init(opaque); |
2057 | if (r) | |
d2308784 | 2058 | goto out_fail; |
cb498ea2 ZX |
2059 | |
2060 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2061 | ||
2062 | if (bad_page == NULL) { | |
2063 | r = -ENOMEM; | |
2064 | goto out; | |
2065 | } | |
2066 | ||
35149e21 AL |
2067 | bad_pfn = page_to_pfn(bad_page); |
2068 | ||
8437a617 | 2069 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2070 | r = -ENOMEM; |
2071 | goto out_free_0; | |
2072 | } | |
2073 | ||
e9b11c17 | 2074 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2075 | if (r < 0) |
7f59f492 | 2076 | goto out_free_0a; |
6aa8b732 | 2077 | |
002c7f7c YS |
2078 | for_each_online_cpu(cpu) { |
2079 | smp_call_function_single(cpu, | |
e9b11c17 | 2080 | kvm_arch_check_processor_compat, |
8691e5a8 | 2081 | &r, 1); |
002c7f7c | 2082 | if (r < 0) |
d2308784 | 2083 | goto out_free_1; |
002c7f7c YS |
2084 | } |
2085 | ||
774c47f1 AK |
2086 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2087 | if (r) | |
d2308784 | 2088 | goto out_free_2; |
6aa8b732 AK |
2089 | register_reboot_notifier(&kvm_reboot_notifier); |
2090 | ||
59ae6c6b AK |
2091 | r = sysdev_class_register(&kvm_sysdev_class); |
2092 | if (r) | |
d2308784 | 2093 | goto out_free_3; |
59ae6c6b AK |
2094 | |
2095 | r = sysdev_register(&kvm_sysdev); | |
2096 | if (r) | |
d2308784 | 2097 | goto out_free_4; |
59ae6c6b | 2098 | |
c16f862d RR |
2099 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
2100 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
2101 | __alignof__(struct kvm_vcpu), |
2102 | 0, NULL); | |
c16f862d RR |
2103 | if (!kvm_vcpu_cache) { |
2104 | r = -ENOMEM; | |
d2308784 | 2105 | goto out_free_5; |
c16f862d RR |
2106 | } |
2107 | ||
6aa8b732 | 2108 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2109 | kvm_vm_fops.owner = module; |
2110 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2111 | |
2112 | r = misc_register(&kvm_dev); | |
2113 | if (r) { | |
d77c26fc | 2114 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2115 | goto out_free; |
2116 | } | |
2117 | ||
15ad7146 AK |
2118 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2119 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2120 | ||
0ea4ed8e DW |
2121 | kvm_init_debug(); |
2122 | ||
c7addb90 | 2123 | return 0; |
6aa8b732 AK |
2124 | |
2125 | out_free: | |
c16f862d | 2126 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2127 | out_free_5: |
59ae6c6b | 2128 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2129 | out_free_4: |
59ae6c6b | 2130 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2131 | out_free_3: |
6aa8b732 | 2132 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2133 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2134 | out_free_2: |
d2308784 | 2135 | out_free_1: |
e9b11c17 | 2136 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2137 | out_free_0a: |
2138 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 ZX |
2139 | out_free_0: |
2140 | __free_page(bad_page); | |
ca45aaae | 2141 | out: |
f8c16bba | 2142 | kvm_arch_exit(); |
d2308784 | 2143 | out_fail: |
6aa8b732 AK |
2144 | return r; |
2145 | } | |
cb498ea2 | 2146 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2147 | |
cb498ea2 | 2148 | void kvm_exit(void) |
6aa8b732 | 2149 | { |
229456fc | 2150 | tracepoint_synchronize_unregister(); |
0ea4ed8e | 2151 | kvm_exit_debug(); |
6aa8b732 | 2152 | misc_deregister(&kvm_dev); |
c16f862d | 2153 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2154 | sysdev_unregister(&kvm_sysdev); |
2155 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2156 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2157 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2158 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2159 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2160 | kvm_arch_exit(); |
7f59f492 | 2161 | free_cpumask_var(cpus_hardware_enabled); |
cea7bb21 | 2162 | __free_page(bad_page); |
6aa8b732 | 2163 | } |
cb498ea2 | 2164 | EXPORT_SYMBOL_GPL(kvm_exit); |