]>
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> |
6aa8b732 | 44 | |
e495606d | 45 | #include <asm/processor.h> |
e495606d AK |
46 | #include <asm/io.h> |
47 | #include <asm/uaccess.h> | |
3e021bf5 | 48 | #include <asm/pgtable.h> |
6aa8b732 AK |
49 | |
50 | MODULE_AUTHOR("Qumranet"); | |
51 | MODULE_LICENSE("GPL"); | |
52 | ||
e9b11c17 ZX |
53 | DEFINE_SPINLOCK(kvm_lock); |
54 | LIST_HEAD(vm_list); | |
133de902 | 55 | |
1b6c0168 AK |
56 | static cpumask_t cpus_hardware_enabled; |
57 | ||
c16f862d RR |
58 | struct kmem_cache *kvm_vcpu_cache; |
59 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 60 | |
15ad7146 AK |
61 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
62 | ||
76f7c879 | 63 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 64 | |
bccf2150 AK |
65 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
66 | unsigned long arg); | |
67 | ||
5aacf0ca JM |
68 | static inline int valid_vcpu(int n) |
69 | { | |
70 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
71 | } | |
72 | ||
bccf2150 AK |
73 | /* |
74 | * Switches to specified vcpu, until a matching vcpu_put() | |
75 | */ | |
313a3dc7 | 76 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 77 | { |
15ad7146 AK |
78 | int cpu; |
79 | ||
bccf2150 | 80 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
81 | cpu = get_cpu(); |
82 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 83 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 84 | put_cpu(); |
6aa8b732 AK |
85 | } |
86 | ||
313a3dc7 | 87 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 88 | { |
15ad7146 | 89 | preempt_disable(); |
313a3dc7 | 90 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
91 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
92 | preempt_enable(); | |
6aa8b732 AK |
93 | mutex_unlock(&vcpu->mutex); |
94 | } | |
95 | ||
d9e368d6 AK |
96 | static void ack_flush(void *_completed) |
97 | { | |
d9e368d6 AK |
98 | } |
99 | ||
100 | void kvm_flush_remote_tlbs(struct kvm *kvm) | |
101 | { | |
49d3bd7e | 102 | int i, cpu; |
d9e368d6 AK |
103 | cpumask_t cpus; |
104 | struct kvm_vcpu *vcpu; | |
d9e368d6 | 105 | |
d9e368d6 | 106 | cpus_clear(cpus); |
fb3f0f51 RR |
107 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
108 | vcpu = kvm->vcpus[i]; | |
109 | if (!vcpu) | |
110 | continue; | |
3176bc3e | 111 | if (test_and_set_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests)) |
d9e368d6 AK |
112 | continue; |
113 | cpu = vcpu->cpu; | |
114 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
49d3bd7e | 115 | cpu_set(cpu, cpus); |
d9e368d6 | 116 | } |
0f74a24c AK |
117 | if (cpus_empty(cpus)) |
118 | return; | |
119 | ++kvm->stat.remote_tlb_flush; | |
49d3bd7e | 120 | smp_call_function_mask(cpus, ack_flush, NULL, 1); |
d9e368d6 AK |
121 | } |
122 | ||
2e53d63a MT |
123 | void kvm_reload_remote_mmus(struct kvm *kvm) |
124 | { | |
125 | int i, cpu; | |
126 | cpumask_t cpus; | |
127 | struct kvm_vcpu *vcpu; | |
128 | ||
129 | cpus_clear(cpus); | |
130 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
131 | vcpu = kvm->vcpus[i]; | |
132 | if (!vcpu) | |
133 | continue; | |
134 | if (test_and_set_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests)) | |
135 | continue; | |
136 | cpu = vcpu->cpu; | |
137 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
138 | cpu_set(cpu, cpus); | |
139 | } | |
140 | if (cpus_empty(cpus)) | |
141 | return; | |
142 | smp_call_function_mask(cpus, ack_flush, NULL, 1); | |
143 | } | |
144 | ||
145 | ||
fb3f0f51 RR |
146 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
147 | { | |
148 | struct page *page; | |
149 | int r; | |
150 | ||
151 | mutex_init(&vcpu->mutex); | |
152 | vcpu->cpu = -1; | |
fb3f0f51 RR |
153 | vcpu->kvm = kvm; |
154 | vcpu->vcpu_id = id; | |
b6958ce4 | 155 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
156 | |
157 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
158 | if (!page) { | |
159 | r = -ENOMEM; | |
160 | goto fail; | |
161 | } | |
162 | vcpu->run = page_address(page); | |
163 | ||
e9b11c17 | 164 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 165 | if (r < 0) |
e9b11c17 | 166 | goto fail_free_run; |
fb3f0f51 RR |
167 | return 0; |
168 | ||
fb3f0f51 RR |
169 | fail_free_run: |
170 | free_page((unsigned long)vcpu->run); | |
171 | fail: | |
76fafa5e | 172 | return r; |
fb3f0f51 RR |
173 | } |
174 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
175 | ||
176 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
177 | { | |
e9b11c17 | 178 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
179 | free_page((unsigned long)vcpu->run); |
180 | } | |
181 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
182 | ||
f17abe9a | 183 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 184 | { |
d19a9cd2 | 185 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 186 | |
d19a9cd2 ZX |
187 | if (IS_ERR(kvm)) |
188 | goto out; | |
6aa8b732 | 189 | |
6d4e4c4f AK |
190 | kvm->mm = current->mm; |
191 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 192 | spin_lock_init(&kvm->mmu_lock); |
74906345 | 193 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 194 | mutex_init(&kvm->lock); |
2eeb2e94 | 195 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 196 | init_rwsem(&kvm->slots_lock); |
d39f13b0 | 197 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
198 | spin_lock(&kvm_lock); |
199 | list_add(&kvm->vm_list, &vm_list); | |
200 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 201 | out: |
f17abe9a AK |
202 | return kvm; |
203 | } | |
204 | ||
6aa8b732 AK |
205 | /* |
206 | * Free any memory in @free but not in @dont. | |
207 | */ | |
208 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
209 | struct kvm_memory_slot *dont) | |
210 | { | |
290fc38d IE |
211 | if (!dont || free->rmap != dont->rmap) |
212 | vfree(free->rmap); | |
6aa8b732 AK |
213 | |
214 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
215 | vfree(free->dirty_bitmap); | |
216 | ||
05da4558 MT |
217 | if (!dont || free->lpage_info != dont->lpage_info) |
218 | vfree(free->lpage_info); | |
219 | ||
6aa8b732 | 220 | free->npages = 0; |
8b6d44c7 | 221 | free->dirty_bitmap = NULL; |
8d4e1288 | 222 | free->rmap = NULL; |
05da4558 | 223 | free->lpage_info = NULL; |
6aa8b732 AK |
224 | } |
225 | ||
d19a9cd2 | 226 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
227 | { |
228 | int i; | |
229 | ||
230 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 231 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
232 | } |
233 | ||
f17abe9a AK |
234 | static void kvm_destroy_vm(struct kvm *kvm) |
235 | { | |
6d4e4c4f AK |
236 | struct mm_struct *mm = kvm->mm; |
237 | ||
133de902 AK |
238 | spin_lock(&kvm_lock); |
239 | list_del(&kvm->vm_list); | |
240 | spin_unlock(&kvm_lock); | |
74906345 | 241 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 242 | kvm_io_bus_destroy(&kvm->mmio_bus); |
d19a9cd2 | 243 | kvm_arch_destroy_vm(kvm); |
6d4e4c4f | 244 | mmdrop(mm); |
f17abe9a AK |
245 | } |
246 | ||
d39f13b0 IE |
247 | void kvm_get_kvm(struct kvm *kvm) |
248 | { | |
249 | atomic_inc(&kvm->users_count); | |
250 | } | |
251 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
252 | ||
253 | void kvm_put_kvm(struct kvm *kvm) | |
254 | { | |
255 | if (atomic_dec_and_test(&kvm->users_count)) | |
256 | kvm_destroy_vm(kvm); | |
257 | } | |
258 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
259 | ||
260 | ||
f17abe9a AK |
261 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
262 | { | |
263 | struct kvm *kvm = filp->private_data; | |
264 | ||
d39f13b0 | 265 | kvm_put_kvm(kvm); |
6aa8b732 AK |
266 | return 0; |
267 | } | |
268 | ||
6aa8b732 AK |
269 | /* |
270 | * Allocate some memory and give it an address in the guest physical address | |
271 | * space. | |
272 | * | |
273 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 274 | * |
10589a46 | 275 | * Must be called holding mmap_sem for write. |
6aa8b732 | 276 | */ |
f78e0e2e SY |
277 | int __kvm_set_memory_region(struct kvm *kvm, |
278 | struct kvm_userspace_memory_region *mem, | |
279 | int user_alloc) | |
6aa8b732 AK |
280 | { |
281 | int r; | |
282 | gfn_t base_gfn; | |
283 | unsigned long npages; | |
284 | unsigned long i; | |
285 | struct kvm_memory_slot *memslot; | |
286 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
287 | |
288 | r = -EINVAL; | |
289 | /* General sanity checks */ | |
290 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
291 | goto out; | |
292 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
293 | goto out; | |
e0d62c7f | 294 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
295 | goto out; |
296 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
297 | goto out; | |
298 | ||
299 | memslot = &kvm->memslots[mem->slot]; | |
300 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
301 | npages = mem->memory_size >> PAGE_SHIFT; | |
302 | ||
303 | if (!npages) | |
304 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
305 | ||
6aa8b732 AK |
306 | new = old = *memslot; |
307 | ||
308 | new.base_gfn = base_gfn; | |
309 | new.npages = npages; | |
310 | new.flags = mem->flags; | |
311 | ||
312 | /* Disallow changing a memory slot's size. */ | |
313 | r = -EINVAL; | |
314 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 315 | goto out_free; |
6aa8b732 AK |
316 | |
317 | /* Check for overlaps */ | |
318 | r = -EEXIST; | |
319 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
320 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
321 | ||
322 | if (s == memslot) | |
323 | continue; | |
324 | if (!((base_gfn + npages <= s->base_gfn) || | |
325 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 326 | goto out_free; |
6aa8b732 | 327 | } |
6aa8b732 | 328 | |
6aa8b732 AK |
329 | /* Free page dirty bitmap if unneeded */ |
330 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 331 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
332 | |
333 | r = -ENOMEM; | |
334 | ||
335 | /* Allocate if a slot is being created */ | |
8d4e1288 | 336 | if (npages && !new.rmap) { |
d77c26fc | 337 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
338 | |
339 | if (!new.rmap) | |
f78e0e2e | 340 | goto out_free; |
290fc38d | 341 | |
290fc38d | 342 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 343 | |
80b14b5b | 344 | new.user_alloc = user_alloc; |
0de10343 | 345 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 346 | } |
05da4558 MT |
347 | if (npages && !new.lpage_info) { |
348 | int largepages = npages / KVM_PAGES_PER_HPAGE; | |
349 | if (npages % KVM_PAGES_PER_HPAGE) | |
350 | largepages++; | |
351 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
352 | largepages++; | |
353 | ||
354 | new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info)); | |
355 | ||
356 | if (!new.lpage_info) | |
357 | goto out_free; | |
358 | ||
359 | memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info)); | |
360 | ||
361 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
362 | new.lpage_info[0].write_count = 1; | |
363 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE) | |
364 | new.lpage_info[largepages-1].write_count = 1; | |
365 | } | |
6aa8b732 AK |
366 | |
367 | /* Allocate page dirty bitmap if needed */ | |
368 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
369 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
370 | ||
371 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
372 | if (!new.dirty_bitmap) | |
f78e0e2e | 373 | goto out_free; |
6aa8b732 AK |
374 | memset(new.dirty_bitmap, 0, dirty_bytes); |
375 | } | |
376 | ||
6aa8b732 AK |
377 | if (mem->slot >= kvm->nmemslots) |
378 | kvm->nmemslots = mem->slot + 1; | |
379 | ||
3ad82a7e ZX |
380 | *memslot = new; |
381 | ||
0de10343 ZX |
382 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
383 | if (r) { | |
384 | *memslot = old; | |
385 | goto out_free; | |
82ce2c96 IE |
386 | } |
387 | ||
6aa8b732 AK |
388 | kvm_free_physmem_slot(&old, &new); |
389 | return 0; | |
390 | ||
f78e0e2e | 391 | out_free: |
6aa8b732 AK |
392 | kvm_free_physmem_slot(&new, &old); |
393 | out: | |
394 | return r; | |
210c7c4d IE |
395 | |
396 | } | |
f78e0e2e SY |
397 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
398 | ||
399 | int kvm_set_memory_region(struct kvm *kvm, | |
400 | struct kvm_userspace_memory_region *mem, | |
401 | int user_alloc) | |
402 | { | |
403 | int r; | |
404 | ||
72dc67a6 | 405 | down_write(&kvm->slots_lock); |
f78e0e2e | 406 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 407 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
408 | return r; |
409 | } | |
210c7c4d IE |
410 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
411 | ||
1fe779f8 CO |
412 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
413 | struct | |
414 | kvm_userspace_memory_region *mem, | |
415 | int user_alloc) | |
210c7c4d | 416 | { |
e0d62c7f IE |
417 | if (mem->slot >= KVM_MEMORY_SLOTS) |
418 | return -EINVAL; | |
210c7c4d | 419 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
420 | } |
421 | ||
5bb064dc ZX |
422 | int kvm_get_dirty_log(struct kvm *kvm, |
423 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
424 | { |
425 | struct kvm_memory_slot *memslot; | |
426 | int r, i; | |
427 | int n; | |
428 | unsigned long any = 0; | |
429 | ||
6aa8b732 AK |
430 | r = -EINVAL; |
431 | if (log->slot >= KVM_MEMORY_SLOTS) | |
432 | goto out; | |
433 | ||
434 | memslot = &kvm->memslots[log->slot]; | |
435 | r = -ENOENT; | |
436 | if (!memslot->dirty_bitmap) | |
437 | goto out; | |
438 | ||
cd1a4a98 | 439 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 440 | |
cd1a4a98 | 441 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
442 | any = memslot->dirty_bitmap[i]; |
443 | ||
444 | r = -EFAULT; | |
445 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
446 | goto out; | |
447 | ||
5bb064dc ZX |
448 | if (any) |
449 | *is_dirty = 1; | |
6aa8b732 AK |
450 | |
451 | r = 0; | |
6aa8b732 | 452 | out: |
6aa8b732 AK |
453 | return r; |
454 | } | |
455 | ||
cea7bb21 IE |
456 | int is_error_page(struct page *page) |
457 | { | |
458 | return page == bad_page; | |
459 | } | |
460 | EXPORT_SYMBOL_GPL(is_error_page); | |
461 | ||
35149e21 AL |
462 | int is_error_pfn(pfn_t pfn) |
463 | { | |
464 | return pfn == bad_pfn; | |
465 | } | |
466 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
467 | ||
f9d46eb0 IE |
468 | static inline unsigned long bad_hva(void) |
469 | { | |
470 | return PAGE_OFFSET; | |
471 | } | |
472 | ||
473 | int kvm_is_error_hva(unsigned long addr) | |
474 | { | |
475 | return addr == bad_hva(); | |
476 | } | |
477 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
478 | ||
e8207547 | 479 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
480 | { |
481 | int i; | |
482 | ||
483 | for (i = 0; i < kvm->nmemslots; ++i) { | |
484 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
485 | ||
486 | if (gfn >= memslot->base_gfn | |
487 | && gfn < memslot->base_gfn + memslot->npages) | |
488 | return memslot; | |
489 | } | |
8b6d44c7 | 490 | return NULL; |
6aa8b732 | 491 | } |
e8207547 AK |
492 | |
493 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
494 | { | |
495 | gfn = unalias_gfn(kvm, gfn); | |
496 | return __gfn_to_memslot(kvm, gfn); | |
497 | } | |
6aa8b732 | 498 | |
e0d62c7f IE |
499 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
500 | { | |
501 | int i; | |
502 | ||
503 | gfn = unalias_gfn(kvm, gfn); | |
504 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
505 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
506 | ||
507 | if (gfn >= memslot->base_gfn | |
508 | && gfn < memslot->base_gfn + memslot->npages) | |
509 | return 1; | |
510 | } | |
511 | return 0; | |
512 | } | |
513 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
514 | ||
05da4558 | 515 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
516 | { |
517 | struct kvm_memory_slot *slot; | |
518 | ||
519 | gfn = unalias_gfn(kvm, gfn); | |
520 | slot = __gfn_to_memslot(kvm, gfn); | |
521 | if (!slot) | |
522 | return bad_hva(); | |
523 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
524 | } | |
525 | ||
aab61cc0 AL |
526 | /* |
527 | * Requires current->mm->mmap_sem to be held | |
528 | */ | |
35149e21 | 529 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 530 | { |
8d4e1288 | 531 | struct page *page[1]; |
539cb660 | 532 | unsigned long addr; |
8d4e1288 | 533 | int npages; |
954bbbc2 | 534 | |
60395224 AK |
535 | might_sleep(); |
536 | ||
539cb660 IE |
537 | addr = gfn_to_hva(kvm, gfn); |
538 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 539 | get_page(bad_page); |
35149e21 | 540 | return page_to_pfn(bad_page); |
8a7ae055 | 541 | } |
8d4e1288 | 542 | |
539cb660 IE |
543 | npages = get_user_pages(current, current->mm, addr, 1, 1, 1, page, |
544 | NULL); | |
545 | ||
8d4e1288 AL |
546 | if (npages != 1) { |
547 | get_page(bad_page); | |
35149e21 | 548 | return page_to_pfn(bad_page); |
8a7ae055 | 549 | } |
8d4e1288 | 550 | |
35149e21 AL |
551 | return page_to_pfn(page[0]); |
552 | } | |
553 | ||
554 | EXPORT_SYMBOL_GPL(gfn_to_pfn); | |
555 | ||
556 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
557 | { | |
558 | return pfn_to_page(gfn_to_pfn(kvm, gfn)); | |
954bbbc2 | 559 | } |
aab61cc0 | 560 | |
954bbbc2 AK |
561 | EXPORT_SYMBOL_GPL(gfn_to_page); |
562 | ||
b4231d61 IE |
563 | void kvm_release_page_clean(struct page *page) |
564 | { | |
35149e21 | 565 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
566 | } |
567 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
568 | ||
35149e21 AL |
569 | void kvm_release_pfn_clean(pfn_t pfn) |
570 | { | |
571 | put_page(pfn_to_page(pfn)); | |
572 | } | |
573 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
574 | ||
b4231d61 | 575 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 576 | { |
35149e21 AL |
577 | kvm_release_pfn_dirty(page_to_pfn(page)); |
578 | } | |
579 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
580 | ||
581 | void kvm_release_pfn_dirty(pfn_t pfn) | |
582 | { | |
583 | kvm_set_pfn_dirty(pfn); | |
584 | kvm_release_pfn_clean(pfn); | |
585 | } | |
586 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
587 | ||
588 | void kvm_set_page_dirty(struct page *page) | |
589 | { | |
590 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
591 | } | |
592 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
593 | ||
594 | void kvm_set_pfn_dirty(pfn_t pfn) | |
595 | { | |
596 | struct page *page = pfn_to_page(pfn); | |
8a7ae055 IE |
597 | if (!PageReserved(page)) |
598 | SetPageDirty(page); | |
8a7ae055 | 599 | } |
35149e21 AL |
600 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
601 | ||
602 | void kvm_set_pfn_accessed(pfn_t pfn) | |
603 | { | |
604 | mark_page_accessed(pfn_to_page(pfn)); | |
605 | } | |
606 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
607 | ||
608 | void kvm_get_pfn(pfn_t pfn) | |
609 | { | |
610 | get_page(pfn_to_page(pfn)); | |
611 | } | |
612 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 613 | |
195aefde IE |
614 | static int next_segment(unsigned long len, int offset) |
615 | { | |
616 | if (len > PAGE_SIZE - offset) | |
617 | return PAGE_SIZE - offset; | |
618 | else | |
619 | return len; | |
620 | } | |
621 | ||
622 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
623 | int len) | |
624 | { | |
e0506bcb IE |
625 | int r; |
626 | unsigned long addr; | |
195aefde | 627 | |
e0506bcb IE |
628 | addr = gfn_to_hva(kvm, gfn); |
629 | if (kvm_is_error_hva(addr)) | |
630 | return -EFAULT; | |
631 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
632 | if (r) | |
195aefde | 633 | return -EFAULT; |
195aefde IE |
634 | return 0; |
635 | } | |
636 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
637 | ||
638 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
639 | { | |
640 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
641 | int seg; | |
642 | int offset = offset_in_page(gpa); | |
643 | int ret; | |
644 | ||
645 | while ((seg = next_segment(len, offset)) != 0) { | |
646 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
647 | if (ret < 0) | |
648 | return ret; | |
649 | offset = 0; | |
650 | len -= seg; | |
651 | data += seg; | |
652 | ++gfn; | |
653 | } | |
654 | return 0; | |
655 | } | |
656 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
657 | ||
7ec54588 MT |
658 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
659 | unsigned long len) | |
660 | { | |
661 | int r; | |
662 | unsigned long addr; | |
663 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
664 | int offset = offset_in_page(gpa); | |
665 | ||
666 | addr = gfn_to_hva(kvm, gfn); | |
667 | if (kvm_is_error_hva(addr)) | |
668 | return -EFAULT; | |
0aac03f0 | 669 | pagefault_disable(); |
7ec54588 | 670 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 671 | pagefault_enable(); |
7ec54588 MT |
672 | if (r) |
673 | return -EFAULT; | |
674 | return 0; | |
675 | } | |
676 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
677 | ||
195aefde IE |
678 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
679 | int offset, int len) | |
680 | { | |
e0506bcb IE |
681 | int r; |
682 | unsigned long addr; | |
195aefde | 683 | |
e0506bcb IE |
684 | addr = gfn_to_hva(kvm, gfn); |
685 | if (kvm_is_error_hva(addr)) | |
686 | return -EFAULT; | |
687 | r = copy_to_user((void __user *)addr + offset, data, len); | |
688 | if (r) | |
195aefde | 689 | return -EFAULT; |
195aefde IE |
690 | mark_page_dirty(kvm, gfn); |
691 | return 0; | |
692 | } | |
693 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
694 | ||
695 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
696 | unsigned long len) | |
697 | { | |
698 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
699 | int seg; | |
700 | int offset = offset_in_page(gpa); | |
701 | int ret; | |
702 | ||
703 | while ((seg = next_segment(len, offset)) != 0) { | |
704 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
705 | if (ret < 0) | |
706 | return ret; | |
707 | offset = 0; | |
708 | len -= seg; | |
709 | data += seg; | |
710 | ++gfn; | |
711 | } | |
712 | return 0; | |
713 | } | |
714 | ||
715 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
716 | { | |
3e021bf5 | 717 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
718 | } |
719 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
720 | ||
721 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
722 | { | |
723 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
724 | int seg; | |
725 | int offset = offset_in_page(gpa); | |
726 | int ret; | |
727 | ||
728 | while ((seg = next_segment(len, offset)) != 0) { | |
729 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
730 | if (ret < 0) | |
731 | return ret; | |
732 | offset = 0; | |
733 | len -= seg; | |
734 | ++gfn; | |
735 | } | |
736 | return 0; | |
737 | } | |
738 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
739 | ||
6aa8b732 AK |
740 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
741 | { | |
31389947 | 742 | struct kvm_memory_slot *memslot; |
6aa8b732 | 743 | |
3b6fff19 | 744 | gfn = unalias_gfn(kvm, gfn); |
7e9d619d RR |
745 | memslot = __gfn_to_memslot(kvm, gfn); |
746 | if (memslot && memslot->dirty_bitmap) { | |
747 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 748 | |
7e9d619d RR |
749 | /* avoid RMW */ |
750 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
751 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
752 | } |
753 | } | |
754 | ||
b6958ce4 ED |
755 | /* |
756 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
757 | */ | |
8776e519 | 758 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 759 | { |
b6958ce4 ED |
760 | DECLARE_WAITQUEUE(wait, current); |
761 | ||
762 | add_wait_queue(&vcpu->wq, &wait); | |
763 | ||
764 | /* | |
765 | * We will block until either an interrupt or a signal wakes us up | |
766 | */ | |
c5ec1534 | 767 | while (!kvm_cpu_has_interrupt(vcpu) |
3d80840d | 768 | && !kvm_cpu_has_pending_timer(vcpu) |
c5ec1534 | 769 | && !signal_pending(current) |
53e0aa7b | 770 | && !kvm_arch_vcpu_runnable(vcpu)) { |
b6958ce4 ED |
771 | set_current_state(TASK_INTERRUPTIBLE); |
772 | vcpu_put(vcpu); | |
773 | schedule(); | |
774 | vcpu_load(vcpu); | |
775 | } | |
d3bef15f | 776 | |
c5ec1534 | 777 | __set_current_state(TASK_RUNNING); |
b6958ce4 | 778 | remove_wait_queue(&vcpu->wq, &wait); |
b6958ce4 ED |
779 | } |
780 | ||
6aa8b732 AK |
781 | void kvm_resched(struct kvm_vcpu *vcpu) |
782 | { | |
3fca0365 YD |
783 | if (!need_resched()) |
784 | return; | |
6aa8b732 | 785 | cond_resched(); |
6aa8b732 AK |
786 | } |
787 | EXPORT_SYMBOL_GPL(kvm_resched); | |
788 | ||
e4a533a4 | 789 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
790 | { |
791 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
792 | struct page *page; |
793 | ||
e4a533a4 | 794 | if (vmf->pgoff == 0) |
039576c0 | 795 | page = virt_to_page(vcpu->run); |
09566765 | 796 | #ifdef CONFIG_X86 |
e4a533a4 | 797 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 798 | page = virt_to_page(vcpu->arch.pio_data); |
09566765 | 799 | #endif |
039576c0 | 800 | else |
e4a533a4 | 801 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 802 | get_page(page); |
e4a533a4 NP |
803 | vmf->page = page; |
804 | return 0; | |
9a2bb7f4 AK |
805 | } |
806 | ||
807 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
e4a533a4 | 808 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
809 | }; |
810 | ||
811 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
812 | { | |
813 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
814 | return 0; | |
815 | } | |
816 | ||
bccf2150 AK |
817 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
818 | { | |
819 | struct kvm_vcpu *vcpu = filp->private_data; | |
820 | ||
66c0b394 | 821 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
822 | return 0; |
823 | } | |
824 | ||
5c502742 | 825 | static const struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
826 | .release = kvm_vcpu_release, |
827 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
828 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 829 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
830 | }; |
831 | ||
832 | /* | |
833 | * Allocates an inode for the vcpu. | |
834 | */ | |
835 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
836 | { | |
837 | int fd, r; | |
838 | struct inode *inode; | |
839 | struct file *file; | |
840 | ||
d6d28168 AK |
841 | r = anon_inode_getfd(&fd, &inode, &file, |
842 | "kvm-vcpu", &kvm_vcpu_fops, vcpu); | |
66c0b394 AV |
843 | if (r) { |
844 | kvm_put_kvm(vcpu->kvm); | |
d6d28168 | 845 | return r; |
66c0b394 | 846 | } |
bccf2150 | 847 | return fd; |
bccf2150 AK |
848 | } |
849 | ||
c5ea7660 AK |
850 | /* |
851 | * Creates some virtual cpus. Good luck creating more than one. | |
852 | */ | |
853 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
854 | { | |
855 | int r; | |
856 | struct kvm_vcpu *vcpu; | |
857 | ||
c5ea7660 | 858 | if (!valid_vcpu(n)) |
fb3f0f51 | 859 | return -EINVAL; |
c5ea7660 | 860 | |
e9b11c17 | 861 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
862 | if (IS_ERR(vcpu)) |
863 | return PTR_ERR(vcpu); | |
c5ea7660 | 864 | |
15ad7146 AK |
865 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
866 | ||
26e5215f AK |
867 | r = kvm_arch_vcpu_setup(vcpu); |
868 | if (r) | |
869 | goto vcpu_destroy; | |
870 | ||
11ec2804 | 871 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
872 | if (kvm->vcpus[n]) { |
873 | r = -EEXIST; | |
11ec2804 | 874 | mutex_unlock(&kvm->lock); |
e9b11c17 | 875 | goto vcpu_destroy; |
fb3f0f51 RR |
876 | } |
877 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 878 | mutex_unlock(&kvm->lock); |
c5ea7660 | 879 | |
fb3f0f51 | 880 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 881 | kvm_get_kvm(kvm); |
bccf2150 AK |
882 | r = create_vcpu_fd(vcpu); |
883 | if (r < 0) | |
fb3f0f51 RR |
884 | goto unlink; |
885 | return r; | |
39c3b86e | 886 | |
fb3f0f51 | 887 | unlink: |
11ec2804 | 888 | mutex_lock(&kvm->lock); |
fb3f0f51 | 889 | kvm->vcpus[n] = NULL; |
11ec2804 | 890 | mutex_unlock(&kvm->lock); |
e9b11c17 | 891 | vcpu_destroy: |
d40ccc62 | 892 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
893 | return r; |
894 | } | |
895 | ||
1961d276 AK |
896 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
897 | { | |
898 | if (sigset) { | |
899 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
900 | vcpu->sigset_active = 1; | |
901 | vcpu->sigset = *sigset; | |
902 | } else | |
903 | vcpu->sigset_active = 0; | |
904 | return 0; | |
905 | } | |
906 | ||
bccf2150 AK |
907 | static long kvm_vcpu_ioctl(struct file *filp, |
908 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 909 | { |
bccf2150 | 910 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 911 | void __user *argp = (void __user *)arg; |
313a3dc7 | 912 | int r; |
6aa8b732 | 913 | |
6d4e4c4f AK |
914 | if (vcpu->kvm->mm != current->mm) |
915 | return -EIO; | |
6aa8b732 | 916 | switch (ioctl) { |
9a2bb7f4 | 917 | case KVM_RUN: |
f0fe5108 AK |
918 | r = -EINVAL; |
919 | if (arg) | |
920 | goto out; | |
b6c7a5dc | 921 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 922 | break; |
6aa8b732 | 923 | case KVM_GET_REGS: { |
3e4bb3ac | 924 | struct kvm_regs *kvm_regs; |
6aa8b732 | 925 | |
3e4bb3ac XZ |
926 | r = -ENOMEM; |
927 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
928 | if (!kvm_regs) | |
6aa8b732 | 929 | goto out; |
3e4bb3ac XZ |
930 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
931 | if (r) | |
932 | goto out_free1; | |
6aa8b732 | 933 | r = -EFAULT; |
3e4bb3ac XZ |
934 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
935 | goto out_free1; | |
6aa8b732 | 936 | r = 0; |
3e4bb3ac XZ |
937 | out_free1: |
938 | kfree(kvm_regs); | |
6aa8b732 AK |
939 | break; |
940 | } | |
941 | case KVM_SET_REGS: { | |
3e4bb3ac | 942 | struct kvm_regs *kvm_regs; |
6aa8b732 | 943 | |
3e4bb3ac XZ |
944 | r = -ENOMEM; |
945 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
946 | if (!kvm_regs) | |
6aa8b732 | 947 | goto out; |
3e4bb3ac XZ |
948 | r = -EFAULT; |
949 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
950 | goto out_free2; | |
951 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 952 | if (r) |
3e4bb3ac | 953 | goto out_free2; |
6aa8b732 | 954 | r = 0; |
3e4bb3ac XZ |
955 | out_free2: |
956 | kfree(kvm_regs); | |
6aa8b732 AK |
957 | break; |
958 | } | |
959 | case KVM_GET_SREGS: { | |
960 | struct kvm_sregs kvm_sregs; | |
961 | ||
bccf2150 | 962 | memset(&kvm_sregs, 0, sizeof kvm_sregs); |
b6c7a5dc | 963 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
964 | if (r) |
965 | goto out; | |
966 | r = -EFAULT; | |
2f366987 | 967 | if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs)) |
6aa8b732 AK |
968 | goto out; |
969 | r = 0; | |
970 | break; | |
971 | } | |
972 | case KVM_SET_SREGS: { | |
973 | struct kvm_sregs kvm_sregs; | |
974 | ||
975 | r = -EFAULT; | |
2f366987 | 976 | if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs)) |
6aa8b732 | 977 | goto out; |
b6c7a5dc | 978 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
979 | if (r) |
980 | goto out; | |
981 | r = 0; | |
982 | break; | |
983 | } | |
62d9f0db MT |
984 | case KVM_GET_MP_STATE: { |
985 | struct kvm_mp_state mp_state; | |
986 | ||
987 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
988 | if (r) | |
989 | goto out; | |
990 | r = -EFAULT; | |
991 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
992 | goto out; | |
993 | r = 0; | |
994 | break; | |
995 | } | |
996 | case KVM_SET_MP_STATE: { | |
997 | struct kvm_mp_state mp_state; | |
998 | ||
999 | r = -EFAULT; | |
1000 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1001 | goto out; | |
1002 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1003 | if (r) | |
1004 | goto out; | |
1005 | r = 0; | |
1006 | break; | |
1007 | } | |
6aa8b732 AK |
1008 | case KVM_TRANSLATE: { |
1009 | struct kvm_translation tr; | |
1010 | ||
1011 | r = -EFAULT; | |
2f366987 | 1012 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1013 | goto out; |
8b006791 | 1014 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1015 | if (r) |
1016 | goto out; | |
1017 | r = -EFAULT; | |
2f366987 | 1018 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1019 | goto out; |
1020 | r = 0; | |
1021 | break; | |
1022 | } | |
6aa8b732 AK |
1023 | case KVM_DEBUG_GUEST: { |
1024 | struct kvm_debug_guest dbg; | |
1025 | ||
1026 | r = -EFAULT; | |
2f366987 | 1027 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1028 | goto out; |
b6c7a5dc | 1029 | r = kvm_arch_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
1030 | if (r) |
1031 | goto out; | |
1032 | r = 0; | |
1033 | break; | |
1034 | } | |
1961d276 AK |
1035 | case KVM_SET_SIGNAL_MASK: { |
1036 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1037 | struct kvm_signal_mask kvm_sigmask; | |
1038 | sigset_t sigset, *p; | |
1039 | ||
1040 | p = NULL; | |
1041 | if (argp) { | |
1042 | r = -EFAULT; | |
1043 | if (copy_from_user(&kvm_sigmask, argp, | |
1044 | sizeof kvm_sigmask)) | |
1045 | goto out; | |
1046 | r = -EINVAL; | |
1047 | if (kvm_sigmask.len != sizeof sigset) | |
1048 | goto out; | |
1049 | r = -EFAULT; | |
1050 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1051 | sizeof sigset)) | |
1052 | goto out; | |
1053 | p = &sigset; | |
1054 | } | |
1055 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1056 | break; | |
1057 | } | |
b8836737 AK |
1058 | case KVM_GET_FPU: { |
1059 | struct kvm_fpu fpu; | |
1060 | ||
1061 | memset(&fpu, 0, sizeof fpu); | |
d0752060 | 1062 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, &fpu); |
b8836737 AK |
1063 | if (r) |
1064 | goto out; | |
1065 | r = -EFAULT; | |
1066 | if (copy_to_user(argp, &fpu, sizeof fpu)) | |
1067 | goto out; | |
1068 | r = 0; | |
1069 | break; | |
1070 | } | |
1071 | case KVM_SET_FPU: { | |
1072 | struct kvm_fpu fpu; | |
1073 | ||
1074 | r = -EFAULT; | |
1075 | if (copy_from_user(&fpu, argp, sizeof fpu)) | |
1076 | goto out; | |
d0752060 | 1077 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, &fpu); |
b8836737 AK |
1078 | if (r) |
1079 | goto out; | |
1080 | r = 0; | |
1081 | break; | |
1082 | } | |
bccf2150 | 1083 | default: |
313a3dc7 | 1084 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1085 | } |
1086 | out: | |
1087 | return r; | |
1088 | } | |
1089 | ||
1090 | static long kvm_vm_ioctl(struct file *filp, | |
1091 | unsigned int ioctl, unsigned long arg) | |
1092 | { | |
1093 | struct kvm *kvm = filp->private_data; | |
1094 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1095 | int r; |
bccf2150 | 1096 | |
6d4e4c4f AK |
1097 | if (kvm->mm != current->mm) |
1098 | return -EIO; | |
bccf2150 AK |
1099 | switch (ioctl) { |
1100 | case KVM_CREATE_VCPU: | |
1101 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1102 | if (r < 0) | |
1103 | goto out; | |
1104 | break; | |
6fc138d2 IE |
1105 | case KVM_SET_USER_MEMORY_REGION: { |
1106 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1107 | ||
1108 | r = -EFAULT; | |
1109 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1110 | sizeof kvm_userspace_mem)) | |
1111 | goto out; | |
1112 | ||
1113 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1114 | if (r) |
1115 | goto out; | |
1116 | break; | |
1117 | } | |
1118 | case KVM_GET_DIRTY_LOG: { | |
1119 | struct kvm_dirty_log log; | |
1120 | ||
1121 | r = -EFAULT; | |
2f366987 | 1122 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1123 | goto out; |
2c6f5df9 | 1124 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1125 | if (r) |
1126 | goto out; | |
1127 | break; | |
1128 | } | |
f17abe9a | 1129 | default: |
1fe779f8 | 1130 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
1131 | } |
1132 | out: | |
1133 | return r; | |
1134 | } | |
1135 | ||
e4a533a4 | 1136 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a AK |
1137 | { |
1138 | struct kvm *kvm = vma->vm_file->private_data; | |
f17abe9a AK |
1139 | struct page *page; |
1140 | ||
e4a533a4 NP |
1141 | if (!kvm_is_visible_gfn(kvm, vmf->pgoff)) |
1142 | return VM_FAULT_SIGBUS; | |
10589a46 | 1143 | page = gfn_to_page(kvm, vmf->pgoff); |
8a7ae055 | 1144 | if (is_error_page(page)) { |
b4231d61 | 1145 | kvm_release_page_clean(page); |
e4a533a4 | 1146 | return VM_FAULT_SIGBUS; |
8a7ae055 | 1147 | } |
e4a533a4 NP |
1148 | vmf->page = page; |
1149 | return 0; | |
f17abe9a AK |
1150 | } |
1151 | ||
1152 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
e4a533a4 | 1153 | .fault = kvm_vm_fault, |
f17abe9a AK |
1154 | }; |
1155 | ||
1156 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1157 | { | |
1158 | vma->vm_ops = &kvm_vm_vm_ops; | |
1159 | return 0; | |
1160 | } | |
1161 | ||
5c502742 | 1162 | static const struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1163 | .release = kvm_vm_release, |
1164 | .unlocked_ioctl = kvm_vm_ioctl, | |
1165 | .compat_ioctl = kvm_vm_ioctl, | |
1166 | .mmap = kvm_vm_mmap, | |
1167 | }; | |
1168 | ||
1169 | static int kvm_dev_ioctl_create_vm(void) | |
1170 | { | |
1171 | int fd, r; | |
1172 | struct inode *inode; | |
1173 | struct file *file; | |
1174 | struct kvm *kvm; | |
1175 | ||
f17abe9a | 1176 | kvm = kvm_create_vm(); |
d6d28168 AK |
1177 | if (IS_ERR(kvm)) |
1178 | return PTR_ERR(kvm); | |
1179 | r = anon_inode_getfd(&fd, &inode, &file, "kvm-vm", &kvm_vm_fops, kvm); | |
1180 | if (r) { | |
66c0b394 | 1181 | kvm_put_kvm(kvm); |
d6d28168 | 1182 | return r; |
f17abe9a AK |
1183 | } |
1184 | ||
f17abe9a | 1185 | return fd; |
f17abe9a AK |
1186 | } |
1187 | ||
1188 | static long kvm_dev_ioctl(struct file *filp, | |
1189 | unsigned int ioctl, unsigned long arg) | |
1190 | { | |
1191 | void __user *argp = (void __user *)arg; | |
07c45a36 | 1192 | long r = -EINVAL; |
f17abe9a AK |
1193 | |
1194 | switch (ioctl) { | |
1195 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1196 | r = -EINVAL; |
1197 | if (arg) | |
1198 | goto out; | |
f17abe9a AK |
1199 | r = KVM_API_VERSION; |
1200 | break; | |
1201 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1202 | r = -EINVAL; |
1203 | if (arg) | |
1204 | goto out; | |
f17abe9a AK |
1205 | r = kvm_dev_ioctl_create_vm(); |
1206 | break; | |
018d00d2 ZX |
1207 | case KVM_CHECK_EXTENSION: |
1208 | r = kvm_dev_ioctl_check_extension((long)argp); | |
5d308f45 | 1209 | break; |
07c45a36 AK |
1210 | case KVM_GET_VCPU_MMAP_SIZE: |
1211 | r = -EINVAL; | |
1212 | if (arg) | |
1213 | goto out; | |
adb1ff46 AK |
1214 | r = PAGE_SIZE; /* struct kvm_run */ |
1215 | #ifdef CONFIG_X86 | |
1216 | r += PAGE_SIZE; /* pio data page */ | |
1217 | #endif | |
07c45a36 | 1218 | break; |
d4c9ff2d FEL |
1219 | case KVM_TRACE_ENABLE: |
1220 | case KVM_TRACE_PAUSE: | |
1221 | case KVM_TRACE_DISABLE: | |
1222 | r = kvm_trace_ioctl(ioctl, arg); | |
1223 | break; | |
6aa8b732 | 1224 | default: |
043405e1 | 1225 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1226 | } |
1227 | out: | |
1228 | return r; | |
1229 | } | |
1230 | ||
6aa8b732 | 1231 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1232 | .unlocked_ioctl = kvm_dev_ioctl, |
1233 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1234 | }; |
1235 | ||
1236 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1237 | KVM_MINOR, |
6aa8b732 AK |
1238 | "kvm", |
1239 | &kvm_chardev_ops, | |
1240 | }; | |
1241 | ||
1b6c0168 AK |
1242 | static void hardware_enable(void *junk) |
1243 | { | |
1244 | int cpu = raw_smp_processor_id(); | |
1245 | ||
1246 | if (cpu_isset(cpu, cpus_hardware_enabled)) | |
1247 | return; | |
1248 | cpu_set(cpu, cpus_hardware_enabled); | |
e9b11c17 | 1249 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
1250 | } |
1251 | ||
1252 | static void hardware_disable(void *junk) | |
1253 | { | |
1254 | int cpu = raw_smp_processor_id(); | |
1255 | ||
1256 | if (!cpu_isset(cpu, cpus_hardware_enabled)) | |
1257 | return; | |
1258 | cpu_clear(cpu, cpus_hardware_enabled); | |
1259 | decache_vcpus_on_cpu(cpu); | |
e9b11c17 | 1260 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1261 | } |
1262 | ||
774c47f1 AK |
1263 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1264 | void *v) | |
1265 | { | |
1266 | int cpu = (long)v; | |
1267 | ||
1a6f4d7f | 1268 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1269 | switch (val) { |
cec9ad27 | 1270 | case CPU_DYING: |
6ec8a856 AK |
1271 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1272 | cpu); | |
1273 | hardware_disable(NULL); | |
1274 | break; | |
774c47f1 | 1275 | case CPU_UP_CANCELED: |
43934a38 JK |
1276 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1277 | cpu); | |
1b6c0168 | 1278 | smp_call_function_single(cpu, hardware_disable, NULL, 0, 1); |
774c47f1 | 1279 | break; |
43934a38 JK |
1280 | case CPU_ONLINE: |
1281 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1282 | cpu); | |
1b6c0168 | 1283 | smp_call_function_single(cpu, hardware_enable, NULL, 0, 1); |
774c47f1 AK |
1284 | break; |
1285 | } | |
1286 | return NOTIFY_OK; | |
1287 | } | |
1288 | ||
9a2b85c6 | 1289 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1290 | void *v) |
9a2b85c6 RR |
1291 | { |
1292 | if (val == SYS_RESTART) { | |
1293 | /* | |
1294 | * Some (well, at least mine) BIOSes hang on reboot if | |
1295 | * in vmx root mode. | |
1296 | */ | |
1297 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1298 | on_each_cpu(hardware_disable, NULL, 0, 1); | |
1299 | } | |
1300 | return NOTIFY_OK; | |
1301 | } | |
1302 | ||
1303 | static struct notifier_block kvm_reboot_notifier = { | |
1304 | .notifier_call = kvm_reboot, | |
1305 | .priority = 0, | |
1306 | }; | |
1307 | ||
2eeb2e94 GH |
1308 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
1309 | { | |
1310 | memset(bus, 0, sizeof(*bus)); | |
1311 | } | |
1312 | ||
1313 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
1314 | { | |
1315 | int i; | |
1316 | ||
1317 | for (i = 0; i < bus->dev_count; i++) { | |
1318 | struct kvm_io_device *pos = bus->devs[i]; | |
1319 | ||
1320 | kvm_iodevice_destructor(pos); | |
1321 | } | |
1322 | } | |
1323 | ||
1324 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr) | |
1325 | { | |
1326 | int i; | |
1327 | ||
1328 | for (i = 0; i < bus->dev_count; i++) { | |
1329 | struct kvm_io_device *pos = bus->devs[i]; | |
1330 | ||
1331 | if (pos->in_range(pos, addr)) | |
1332 | return pos; | |
1333 | } | |
1334 | ||
1335 | return NULL; | |
1336 | } | |
1337 | ||
1338 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
1339 | { | |
1340 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
1341 | ||
1342 | bus->devs[bus->dev_count++] = dev; | |
1343 | } | |
1344 | ||
774c47f1 AK |
1345 | static struct notifier_block kvm_cpu_notifier = { |
1346 | .notifier_call = kvm_cpu_hotplug, | |
1347 | .priority = 20, /* must be > scheduler priority */ | |
1348 | }; | |
1349 | ||
8b88b099 | 1350 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
1351 | { |
1352 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
1353 | struct kvm *kvm; |
1354 | ||
8b88b099 | 1355 | *val = 0; |
ba1389b7 AK |
1356 | spin_lock(&kvm_lock); |
1357 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 1358 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 1359 | spin_unlock(&kvm_lock); |
8b88b099 | 1360 | return 0; |
ba1389b7 AK |
1361 | } |
1362 | ||
1363 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
1364 | ||
8b88b099 | 1365 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
1366 | { |
1367 | unsigned offset = (long)_offset; | |
1165f5fe AK |
1368 | struct kvm *kvm; |
1369 | struct kvm_vcpu *vcpu; | |
1370 | int i; | |
1371 | ||
8b88b099 | 1372 | *val = 0; |
1165f5fe AK |
1373 | spin_lock(&kvm_lock); |
1374 | list_for_each_entry(kvm, &vm_list, vm_list) | |
1375 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
1376 | vcpu = kvm->vcpus[i]; |
1377 | if (vcpu) | |
8b88b099 | 1378 | *val += *(u32 *)((void *)vcpu + offset); |
1165f5fe AK |
1379 | } |
1380 | spin_unlock(&kvm_lock); | |
8b88b099 | 1381 | return 0; |
1165f5fe AK |
1382 | } |
1383 | ||
ba1389b7 AK |
1384 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
1385 | ||
1386 | static struct file_operations *stat_fops[] = { | |
1387 | [KVM_STAT_VCPU] = &vcpu_stat_fops, | |
1388 | [KVM_STAT_VM] = &vm_stat_fops, | |
1389 | }; | |
1165f5fe | 1390 | |
a16b043c | 1391 | static void kvm_init_debug(void) |
6aa8b732 AK |
1392 | { |
1393 | struct kvm_stats_debugfs_item *p; | |
1394 | ||
76f7c879 | 1395 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 1396 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 1397 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 1398 | (void *)(long)p->offset, |
ba1389b7 | 1399 | stat_fops[p->kind]); |
6aa8b732 AK |
1400 | } |
1401 | ||
1402 | static void kvm_exit_debug(void) | |
1403 | { | |
1404 | struct kvm_stats_debugfs_item *p; | |
1405 | ||
1406 | for (p = debugfs_entries; p->name; ++p) | |
1407 | debugfs_remove(p->dentry); | |
76f7c879 | 1408 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
1409 | } |
1410 | ||
59ae6c6b AK |
1411 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
1412 | { | |
4267c41a | 1413 | hardware_disable(NULL); |
59ae6c6b AK |
1414 | return 0; |
1415 | } | |
1416 | ||
1417 | static int kvm_resume(struct sys_device *dev) | |
1418 | { | |
4267c41a | 1419 | hardware_enable(NULL); |
59ae6c6b AK |
1420 | return 0; |
1421 | } | |
1422 | ||
1423 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 1424 | .name = "kvm", |
59ae6c6b AK |
1425 | .suspend = kvm_suspend, |
1426 | .resume = kvm_resume, | |
1427 | }; | |
1428 | ||
1429 | static struct sys_device kvm_sysdev = { | |
1430 | .id = 0, | |
1431 | .cls = &kvm_sysdev_class, | |
1432 | }; | |
1433 | ||
cea7bb21 | 1434 | struct page *bad_page; |
35149e21 | 1435 | pfn_t bad_pfn; |
6aa8b732 | 1436 | |
15ad7146 AK |
1437 | static inline |
1438 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
1439 | { | |
1440 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
1441 | } | |
1442 | ||
1443 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
1444 | { | |
1445 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1446 | ||
e9b11c17 | 1447 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
1448 | } |
1449 | ||
1450 | static void kvm_sched_out(struct preempt_notifier *pn, | |
1451 | struct task_struct *next) | |
1452 | { | |
1453 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1454 | ||
e9b11c17 | 1455 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
1456 | } |
1457 | ||
f8c16bba | 1458 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 1459 | struct module *module) |
6aa8b732 AK |
1460 | { |
1461 | int r; | |
002c7f7c | 1462 | int cpu; |
6aa8b732 | 1463 | |
cb498ea2 ZX |
1464 | kvm_init_debug(); |
1465 | ||
f8c16bba ZX |
1466 | r = kvm_arch_init(opaque); |
1467 | if (r) | |
d2308784 | 1468 | goto out_fail; |
cb498ea2 ZX |
1469 | |
1470 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
1471 | ||
1472 | if (bad_page == NULL) { | |
1473 | r = -ENOMEM; | |
1474 | goto out; | |
1475 | } | |
1476 | ||
35149e21 AL |
1477 | bad_pfn = page_to_pfn(bad_page); |
1478 | ||
e9b11c17 | 1479 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 1480 | if (r < 0) |
d2308784 | 1481 | goto out_free_0; |
6aa8b732 | 1482 | |
002c7f7c YS |
1483 | for_each_online_cpu(cpu) { |
1484 | smp_call_function_single(cpu, | |
e9b11c17 | 1485 | kvm_arch_check_processor_compat, |
002c7f7c YS |
1486 | &r, 0, 1); |
1487 | if (r < 0) | |
d2308784 | 1488 | goto out_free_1; |
002c7f7c YS |
1489 | } |
1490 | ||
1b6c0168 | 1491 | on_each_cpu(hardware_enable, NULL, 0, 1); |
774c47f1 AK |
1492 | r = register_cpu_notifier(&kvm_cpu_notifier); |
1493 | if (r) | |
d2308784 | 1494 | goto out_free_2; |
6aa8b732 AK |
1495 | register_reboot_notifier(&kvm_reboot_notifier); |
1496 | ||
59ae6c6b AK |
1497 | r = sysdev_class_register(&kvm_sysdev_class); |
1498 | if (r) | |
d2308784 | 1499 | goto out_free_3; |
59ae6c6b AK |
1500 | |
1501 | r = sysdev_register(&kvm_sysdev); | |
1502 | if (r) | |
d2308784 | 1503 | goto out_free_4; |
59ae6c6b | 1504 | |
c16f862d RR |
1505 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
1506 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
1507 | __alignof__(struct kvm_vcpu), |
1508 | 0, NULL); | |
c16f862d RR |
1509 | if (!kvm_vcpu_cache) { |
1510 | r = -ENOMEM; | |
d2308784 | 1511 | goto out_free_5; |
c16f862d RR |
1512 | } |
1513 | ||
6aa8b732 AK |
1514 | kvm_chardev_ops.owner = module; |
1515 | ||
1516 | r = misc_register(&kvm_dev); | |
1517 | if (r) { | |
d77c26fc | 1518 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
1519 | goto out_free; |
1520 | } | |
1521 | ||
15ad7146 AK |
1522 | kvm_preempt_ops.sched_in = kvm_sched_in; |
1523 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
1524 | ||
c7addb90 | 1525 | return 0; |
6aa8b732 AK |
1526 | |
1527 | out_free: | |
c16f862d | 1528 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 1529 | out_free_5: |
59ae6c6b | 1530 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 1531 | out_free_4: |
59ae6c6b | 1532 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 1533 | out_free_3: |
6aa8b732 | 1534 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 1535 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 1536 | out_free_2: |
1b6c0168 | 1537 | on_each_cpu(hardware_disable, NULL, 0, 1); |
d2308784 | 1538 | out_free_1: |
e9b11c17 | 1539 | kvm_arch_hardware_unsetup(); |
d2308784 ZX |
1540 | out_free_0: |
1541 | __free_page(bad_page); | |
ca45aaae | 1542 | out: |
f8c16bba | 1543 | kvm_arch_exit(); |
cb498ea2 | 1544 | kvm_exit_debug(); |
d2308784 | 1545 | out_fail: |
6aa8b732 AK |
1546 | return r; |
1547 | } | |
cb498ea2 | 1548 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 1549 | |
cb498ea2 | 1550 | void kvm_exit(void) |
6aa8b732 | 1551 | { |
d4c9ff2d | 1552 | kvm_trace_cleanup(); |
6aa8b732 | 1553 | misc_deregister(&kvm_dev); |
c16f862d | 1554 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
1555 | sysdev_unregister(&kvm_sysdev); |
1556 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 1557 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 1558 | unregister_cpu_notifier(&kvm_cpu_notifier); |
1b6c0168 | 1559 | on_each_cpu(hardware_disable, NULL, 0, 1); |
e9b11c17 | 1560 | kvm_arch_hardware_unsetup(); |
f8c16bba | 1561 | kvm_arch_exit(); |
6aa8b732 | 1562 | kvm_exit_debug(); |
cea7bb21 | 1563 | __free_page(bad_page); |
6aa8b732 | 1564 | } |
cb498ea2 | 1565 | EXPORT_SYMBOL_GPL(kvm_exit); |