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 | ||
4ecac3fd AK |
68 | bool kvm_rebooting; |
69 | ||
5aacf0ca JM |
70 | static inline int valid_vcpu(int n) |
71 | { | |
72 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
73 | } | |
74 | ||
bccf2150 AK |
75 | /* |
76 | * Switches to specified vcpu, until a matching vcpu_put() | |
77 | */ | |
313a3dc7 | 78 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 79 | { |
15ad7146 AK |
80 | int cpu; |
81 | ||
bccf2150 | 82 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
83 | cpu = get_cpu(); |
84 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 85 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 86 | put_cpu(); |
6aa8b732 AK |
87 | } |
88 | ||
313a3dc7 | 89 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 90 | { |
15ad7146 | 91 | preempt_disable(); |
313a3dc7 | 92 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
93 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
94 | preempt_enable(); | |
6aa8b732 AK |
95 | mutex_unlock(&vcpu->mutex); |
96 | } | |
97 | ||
d9e368d6 AK |
98 | static void ack_flush(void *_completed) |
99 | { | |
d9e368d6 AK |
100 | } |
101 | ||
102 | void kvm_flush_remote_tlbs(struct kvm *kvm) | |
103 | { | |
49d3bd7e | 104 | int i, cpu; |
d9e368d6 AK |
105 | cpumask_t cpus; |
106 | struct kvm_vcpu *vcpu; | |
d9e368d6 | 107 | |
d9e368d6 | 108 | cpus_clear(cpus); |
fb3f0f51 RR |
109 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
110 | vcpu = kvm->vcpus[i]; | |
111 | if (!vcpu) | |
112 | continue; | |
3176bc3e | 113 | if (test_and_set_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests)) |
d9e368d6 AK |
114 | continue; |
115 | cpu = vcpu->cpu; | |
116 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
49d3bd7e | 117 | cpu_set(cpu, cpus); |
d9e368d6 | 118 | } |
0f74a24c AK |
119 | if (cpus_empty(cpus)) |
120 | return; | |
121 | ++kvm->stat.remote_tlb_flush; | |
49d3bd7e | 122 | smp_call_function_mask(cpus, ack_flush, NULL, 1); |
d9e368d6 AK |
123 | } |
124 | ||
2e53d63a MT |
125 | void kvm_reload_remote_mmus(struct kvm *kvm) |
126 | { | |
127 | int i, cpu; | |
128 | cpumask_t cpus; | |
129 | struct kvm_vcpu *vcpu; | |
130 | ||
131 | cpus_clear(cpus); | |
132 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
133 | vcpu = kvm->vcpus[i]; | |
134 | if (!vcpu) | |
135 | continue; | |
136 | if (test_and_set_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests)) | |
137 | continue; | |
138 | cpu = vcpu->cpu; | |
139 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
140 | cpu_set(cpu, cpus); | |
141 | } | |
142 | if (cpus_empty(cpus)) | |
143 | return; | |
144 | smp_call_function_mask(cpus, ack_flush, NULL, 1); | |
145 | } | |
146 | ||
147 | ||
fb3f0f51 RR |
148 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
149 | { | |
150 | struct page *page; | |
151 | int r; | |
152 | ||
153 | mutex_init(&vcpu->mutex); | |
154 | vcpu->cpu = -1; | |
fb3f0f51 RR |
155 | vcpu->kvm = kvm; |
156 | vcpu->vcpu_id = id; | |
b6958ce4 | 157 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
158 | |
159 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
160 | if (!page) { | |
161 | r = -ENOMEM; | |
162 | goto fail; | |
163 | } | |
164 | vcpu->run = page_address(page); | |
165 | ||
e9b11c17 | 166 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 167 | if (r < 0) |
e9b11c17 | 168 | goto fail_free_run; |
fb3f0f51 RR |
169 | return 0; |
170 | ||
fb3f0f51 RR |
171 | fail_free_run: |
172 | free_page((unsigned long)vcpu->run); | |
173 | fail: | |
76fafa5e | 174 | return r; |
fb3f0f51 RR |
175 | } |
176 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
177 | ||
178 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
179 | { | |
e9b11c17 | 180 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
181 | free_page((unsigned long)vcpu->run); |
182 | } | |
183 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
184 | ||
f17abe9a | 185 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 186 | { |
d19a9cd2 | 187 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 188 | |
d19a9cd2 ZX |
189 | if (IS_ERR(kvm)) |
190 | goto out; | |
6aa8b732 | 191 | |
6d4e4c4f AK |
192 | kvm->mm = current->mm; |
193 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 194 | spin_lock_init(&kvm->mmu_lock); |
74906345 | 195 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 196 | mutex_init(&kvm->lock); |
2eeb2e94 | 197 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 198 | init_rwsem(&kvm->slots_lock); |
d39f13b0 | 199 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
200 | spin_lock(&kvm_lock); |
201 | list_add(&kvm->vm_list, &vm_list); | |
202 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 203 | out: |
f17abe9a AK |
204 | return kvm; |
205 | } | |
206 | ||
6aa8b732 AK |
207 | /* |
208 | * Free any memory in @free but not in @dont. | |
209 | */ | |
210 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
211 | struct kvm_memory_slot *dont) | |
212 | { | |
290fc38d IE |
213 | if (!dont || free->rmap != dont->rmap) |
214 | vfree(free->rmap); | |
6aa8b732 AK |
215 | |
216 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
217 | vfree(free->dirty_bitmap); | |
218 | ||
05da4558 MT |
219 | if (!dont || free->lpage_info != dont->lpage_info) |
220 | vfree(free->lpage_info); | |
221 | ||
6aa8b732 | 222 | free->npages = 0; |
8b6d44c7 | 223 | free->dirty_bitmap = NULL; |
8d4e1288 | 224 | free->rmap = NULL; |
05da4558 | 225 | free->lpage_info = NULL; |
6aa8b732 AK |
226 | } |
227 | ||
d19a9cd2 | 228 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
229 | { |
230 | int i; | |
231 | ||
232 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 233 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
234 | } |
235 | ||
f17abe9a AK |
236 | static void kvm_destroy_vm(struct kvm *kvm) |
237 | { | |
6d4e4c4f AK |
238 | struct mm_struct *mm = kvm->mm; |
239 | ||
133de902 AK |
240 | spin_lock(&kvm_lock); |
241 | list_del(&kvm->vm_list); | |
242 | spin_unlock(&kvm_lock); | |
74906345 | 243 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 244 | kvm_io_bus_destroy(&kvm->mmio_bus); |
d19a9cd2 | 245 | kvm_arch_destroy_vm(kvm); |
6d4e4c4f | 246 | mmdrop(mm); |
f17abe9a AK |
247 | } |
248 | ||
d39f13b0 IE |
249 | void kvm_get_kvm(struct kvm *kvm) |
250 | { | |
251 | atomic_inc(&kvm->users_count); | |
252 | } | |
253 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
254 | ||
255 | void kvm_put_kvm(struct kvm *kvm) | |
256 | { | |
257 | if (atomic_dec_and_test(&kvm->users_count)) | |
258 | kvm_destroy_vm(kvm); | |
259 | } | |
260 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
261 | ||
262 | ||
f17abe9a AK |
263 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
264 | { | |
265 | struct kvm *kvm = filp->private_data; | |
266 | ||
d39f13b0 | 267 | kvm_put_kvm(kvm); |
6aa8b732 AK |
268 | return 0; |
269 | } | |
270 | ||
6aa8b732 AK |
271 | /* |
272 | * Allocate some memory and give it an address in the guest physical address | |
273 | * space. | |
274 | * | |
275 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 276 | * |
10589a46 | 277 | * Must be called holding mmap_sem for write. |
6aa8b732 | 278 | */ |
f78e0e2e SY |
279 | int __kvm_set_memory_region(struct kvm *kvm, |
280 | struct kvm_userspace_memory_region *mem, | |
281 | int user_alloc) | |
6aa8b732 AK |
282 | { |
283 | int r; | |
284 | gfn_t base_gfn; | |
285 | unsigned long npages; | |
286 | unsigned long i; | |
287 | struct kvm_memory_slot *memslot; | |
288 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
289 | |
290 | r = -EINVAL; | |
291 | /* General sanity checks */ | |
292 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
293 | goto out; | |
294 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
295 | goto out; | |
e0d62c7f | 296 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
297 | goto out; |
298 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
299 | goto out; | |
300 | ||
301 | memslot = &kvm->memslots[mem->slot]; | |
302 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
303 | npages = mem->memory_size >> PAGE_SHIFT; | |
304 | ||
305 | if (!npages) | |
306 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
307 | ||
6aa8b732 AK |
308 | new = old = *memslot; |
309 | ||
310 | new.base_gfn = base_gfn; | |
311 | new.npages = npages; | |
312 | new.flags = mem->flags; | |
313 | ||
314 | /* Disallow changing a memory slot's size. */ | |
315 | r = -EINVAL; | |
316 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 317 | goto out_free; |
6aa8b732 AK |
318 | |
319 | /* Check for overlaps */ | |
320 | r = -EEXIST; | |
321 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
322 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
323 | ||
324 | if (s == memslot) | |
325 | continue; | |
326 | if (!((base_gfn + npages <= s->base_gfn) || | |
327 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 328 | goto out_free; |
6aa8b732 | 329 | } |
6aa8b732 | 330 | |
6aa8b732 AK |
331 | /* Free page dirty bitmap if unneeded */ |
332 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 333 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
334 | |
335 | r = -ENOMEM; | |
336 | ||
337 | /* Allocate if a slot is being created */ | |
8d4e1288 | 338 | if (npages && !new.rmap) { |
d77c26fc | 339 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
340 | |
341 | if (!new.rmap) | |
f78e0e2e | 342 | goto out_free; |
290fc38d | 343 | |
290fc38d | 344 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 345 | |
80b14b5b | 346 | new.user_alloc = user_alloc; |
0de10343 | 347 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 348 | } |
05da4558 MT |
349 | if (npages && !new.lpage_info) { |
350 | int largepages = npages / KVM_PAGES_PER_HPAGE; | |
351 | if (npages % KVM_PAGES_PER_HPAGE) | |
352 | largepages++; | |
353 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
354 | largepages++; | |
355 | ||
356 | new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info)); | |
357 | ||
358 | if (!new.lpage_info) | |
359 | goto out_free; | |
360 | ||
361 | memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info)); | |
362 | ||
363 | if (base_gfn % KVM_PAGES_PER_HPAGE) | |
364 | new.lpage_info[0].write_count = 1; | |
365 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE) | |
366 | new.lpage_info[largepages-1].write_count = 1; | |
367 | } | |
6aa8b732 AK |
368 | |
369 | /* Allocate page dirty bitmap if needed */ | |
370 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
371 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
372 | ||
373 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
374 | if (!new.dirty_bitmap) | |
f78e0e2e | 375 | goto out_free; |
6aa8b732 AK |
376 | memset(new.dirty_bitmap, 0, dirty_bytes); |
377 | } | |
378 | ||
6aa8b732 AK |
379 | if (mem->slot >= kvm->nmemslots) |
380 | kvm->nmemslots = mem->slot + 1; | |
381 | ||
3ad82a7e ZX |
382 | *memslot = new; |
383 | ||
0de10343 ZX |
384 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
385 | if (r) { | |
386 | *memslot = old; | |
387 | goto out_free; | |
82ce2c96 IE |
388 | } |
389 | ||
6aa8b732 AK |
390 | kvm_free_physmem_slot(&old, &new); |
391 | return 0; | |
392 | ||
f78e0e2e | 393 | out_free: |
6aa8b732 AK |
394 | kvm_free_physmem_slot(&new, &old); |
395 | out: | |
396 | return r; | |
210c7c4d IE |
397 | |
398 | } | |
f78e0e2e SY |
399 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
400 | ||
401 | int kvm_set_memory_region(struct kvm *kvm, | |
402 | struct kvm_userspace_memory_region *mem, | |
403 | int user_alloc) | |
404 | { | |
405 | int r; | |
406 | ||
72dc67a6 | 407 | down_write(&kvm->slots_lock); |
f78e0e2e | 408 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 409 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
410 | return r; |
411 | } | |
210c7c4d IE |
412 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
413 | ||
1fe779f8 CO |
414 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
415 | struct | |
416 | kvm_userspace_memory_region *mem, | |
417 | int user_alloc) | |
210c7c4d | 418 | { |
e0d62c7f IE |
419 | if (mem->slot >= KVM_MEMORY_SLOTS) |
420 | return -EINVAL; | |
210c7c4d | 421 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
422 | } |
423 | ||
5bb064dc ZX |
424 | int kvm_get_dirty_log(struct kvm *kvm, |
425 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
426 | { |
427 | struct kvm_memory_slot *memslot; | |
428 | int r, i; | |
429 | int n; | |
430 | unsigned long any = 0; | |
431 | ||
6aa8b732 AK |
432 | r = -EINVAL; |
433 | if (log->slot >= KVM_MEMORY_SLOTS) | |
434 | goto out; | |
435 | ||
436 | memslot = &kvm->memslots[log->slot]; | |
437 | r = -ENOENT; | |
438 | if (!memslot->dirty_bitmap) | |
439 | goto out; | |
440 | ||
cd1a4a98 | 441 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 442 | |
cd1a4a98 | 443 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
444 | any = memslot->dirty_bitmap[i]; |
445 | ||
446 | r = -EFAULT; | |
447 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
448 | goto out; | |
449 | ||
5bb064dc ZX |
450 | if (any) |
451 | *is_dirty = 1; | |
6aa8b732 AK |
452 | |
453 | r = 0; | |
6aa8b732 | 454 | out: |
6aa8b732 AK |
455 | return r; |
456 | } | |
457 | ||
cea7bb21 IE |
458 | int is_error_page(struct page *page) |
459 | { | |
460 | return page == bad_page; | |
461 | } | |
462 | EXPORT_SYMBOL_GPL(is_error_page); | |
463 | ||
35149e21 AL |
464 | int is_error_pfn(pfn_t pfn) |
465 | { | |
466 | return pfn == bad_pfn; | |
467 | } | |
468 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
469 | ||
f9d46eb0 IE |
470 | static inline unsigned long bad_hva(void) |
471 | { | |
472 | return PAGE_OFFSET; | |
473 | } | |
474 | ||
475 | int kvm_is_error_hva(unsigned long addr) | |
476 | { | |
477 | return addr == bad_hva(); | |
478 | } | |
479 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
480 | ||
e8207547 | 481 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
482 | { |
483 | int i; | |
484 | ||
485 | for (i = 0; i < kvm->nmemslots; ++i) { | |
486 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
487 | ||
488 | if (gfn >= memslot->base_gfn | |
489 | && gfn < memslot->base_gfn + memslot->npages) | |
490 | return memslot; | |
491 | } | |
8b6d44c7 | 492 | return NULL; |
6aa8b732 | 493 | } |
e8207547 AK |
494 | |
495 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
496 | { | |
497 | gfn = unalias_gfn(kvm, gfn); | |
498 | return __gfn_to_memslot(kvm, gfn); | |
499 | } | |
6aa8b732 | 500 | |
e0d62c7f IE |
501 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
502 | { | |
503 | int i; | |
504 | ||
505 | gfn = unalias_gfn(kvm, gfn); | |
506 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
507 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
508 | ||
509 | if (gfn >= memslot->base_gfn | |
510 | && gfn < memslot->base_gfn + memslot->npages) | |
511 | return 1; | |
512 | } | |
513 | return 0; | |
514 | } | |
515 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
516 | ||
05da4558 | 517 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
518 | { |
519 | struct kvm_memory_slot *slot; | |
520 | ||
521 | gfn = unalias_gfn(kvm, gfn); | |
522 | slot = __gfn_to_memslot(kvm, gfn); | |
523 | if (!slot) | |
524 | return bad_hva(); | |
525 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
526 | } | |
0d150298 | 527 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 528 | |
aab61cc0 AL |
529 | /* |
530 | * Requires current->mm->mmap_sem to be held | |
531 | */ | |
35149e21 | 532 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 533 | { |
8d4e1288 | 534 | struct page *page[1]; |
539cb660 | 535 | unsigned long addr; |
8d4e1288 | 536 | int npages; |
2e2e3738 | 537 | pfn_t pfn; |
954bbbc2 | 538 | |
60395224 AK |
539 | might_sleep(); |
540 | ||
539cb660 IE |
541 | addr = gfn_to_hva(kvm, gfn); |
542 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 543 | get_page(bad_page); |
35149e21 | 544 | return page_to_pfn(bad_page); |
8a7ae055 | 545 | } |
8d4e1288 | 546 | |
539cb660 IE |
547 | npages = get_user_pages(current, current->mm, addr, 1, 1, 1, page, |
548 | NULL); | |
549 | ||
2e2e3738 AL |
550 | if (unlikely(npages != 1)) { |
551 | struct vm_area_struct *vma; | |
552 | ||
553 | vma = find_vma(current->mm, addr); | |
554 | if (vma == NULL || addr < vma->vm_start || | |
555 | !(vma->vm_flags & VM_PFNMAP)) { | |
556 | get_page(bad_page); | |
557 | return page_to_pfn(bad_page); | |
558 | } | |
559 | ||
560 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
561 | BUG_ON(pfn_valid(pfn)); | |
562 | } else | |
563 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 564 | |
2e2e3738 | 565 | return pfn; |
35149e21 AL |
566 | } |
567 | ||
568 | EXPORT_SYMBOL_GPL(gfn_to_pfn); | |
569 | ||
570 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
571 | { | |
2e2e3738 AL |
572 | pfn_t pfn; |
573 | ||
574 | pfn = gfn_to_pfn(kvm, gfn); | |
575 | if (pfn_valid(pfn)) | |
576 | return pfn_to_page(pfn); | |
577 | ||
578 | WARN_ON(!pfn_valid(pfn)); | |
579 | ||
580 | get_page(bad_page); | |
581 | return bad_page; | |
954bbbc2 | 582 | } |
aab61cc0 | 583 | |
954bbbc2 AK |
584 | EXPORT_SYMBOL_GPL(gfn_to_page); |
585 | ||
b4231d61 IE |
586 | void kvm_release_page_clean(struct page *page) |
587 | { | |
35149e21 | 588 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
589 | } |
590 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
591 | ||
35149e21 AL |
592 | void kvm_release_pfn_clean(pfn_t pfn) |
593 | { | |
2e2e3738 AL |
594 | if (pfn_valid(pfn)) |
595 | put_page(pfn_to_page(pfn)); | |
35149e21 AL |
596 | } |
597 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
598 | ||
b4231d61 | 599 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 600 | { |
35149e21 AL |
601 | kvm_release_pfn_dirty(page_to_pfn(page)); |
602 | } | |
603 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
604 | ||
605 | void kvm_release_pfn_dirty(pfn_t pfn) | |
606 | { | |
607 | kvm_set_pfn_dirty(pfn); | |
608 | kvm_release_pfn_clean(pfn); | |
609 | } | |
610 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
611 | ||
612 | void kvm_set_page_dirty(struct page *page) | |
613 | { | |
614 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
615 | } | |
616 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
617 | ||
618 | void kvm_set_pfn_dirty(pfn_t pfn) | |
619 | { | |
2e2e3738 AL |
620 | if (pfn_valid(pfn)) { |
621 | struct page *page = pfn_to_page(pfn); | |
622 | if (!PageReserved(page)) | |
623 | SetPageDirty(page); | |
624 | } | |
8a7ae055 | 625 | } |
35149e21 AL |
626 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
627 | ||
628 | void kvm_set_pfn_accessed(pfn_t pfn) | |
629 | { | |
2e2e3738 AL |
630 | if (pfn_valid(pfn)) |
631 | mark_page_accessed(pfn_to_page(pfn)); | |
35149e21 AL |
632 | } |
633 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
634 | ||
635 | void kvm_get_pfn(pfn_t pfn) | |
636 | { | |
2e2e3738 AL |
637 | if (pfn_valid(pfn)) |
638 | get_page(pfn_to_page(pfn)); | |
35149e21 AL |
639 | } |
640 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 641 | |
195aefde IE |
642 | static int next_segment(unsigned long len, int offset) |
643 | { | |
644 | if (len > PAGE_SIZE - offset) | |
645 | return PAGE_SIZE - offset; | |
646 | else | |
647 | return len; | |
648 | } | |
649 | ||
650 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
651 | int len) | |
652 | { | |
e0506bcb IE |
653 | int r; |
654 | unsigned long addr; | |
195aefde | 655 | |
e0506bcb IE |
656 | addr = gfn_to_hva(kvm, gfn); |
657 | if (kvm_is_error_hva(addr)) | |
658 | return -EFAULT; | |
659 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
660 | if (r) | |
195aefde | 661 | return -EFAULT; |
195aefde IE |
662 | return 0; |
663 | } | |
664 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
665 | ||
666 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
667 | { | |
668 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
669 | int seg; | |
670 | int offset = offset_in_page(gpa); | |
671 | int ret; | |
672 | ||
673 | while ((seg = next_segment(len, offset)) != 0) { | |
674 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
675 | if (ret < 0) | |
676 | return ret; | |
677 | offset = 0; | |
678 | len -= seg; | |
679 | data += seg; | |
680 | ++gfn; | |
681 | } | |
682 | return 0; | |
683 | } | |
684 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
685 | ||
7ec54588 MT |
686 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
687 | unsigned long len) | |
688 | { | |
689 | int r; | |
690 | unsigned long addr; | |
691 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
692 | int offset = offset_in_page(gpa); | |
693 | ||
694 | addr = gfn_to_hva(kvm, gfn); | |
695 | if (kvm_is_error_hva(addr)) | |
696 | return -EFAULT; | |
0aac03f0 | 697 | pagefault_disable(); |
7ec54588 | 698 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 699 | pagefault_enable(); |
7ec54588 MT |
700 | if (r) |
701 | return -EFAULT; | |
702 | return 0; | |
703 | } | |
704 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
705 | ||
195aefde IE |
706 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
707 | int offset, int len) | |
708 | { | |
e0506bcb IE |
709 | int r; |
710 | unsigned long addr; | |
195aefde | 711 | |
e0506bcb IE |
712 | addr = gfn_to_hva(kvm, gfn); |
713 | if (kvm_is_error_hva(addr)) | |
714 | return -EFAULT; | |
715 | r = copy_to_user((void __user *)addr + offset, data, len); | |
716 | if (r) | |
195aefde | 717 | return -EFAULT; |
195aefde IE |
718 | mark_page_dirty(kvm, gfn); |
719 | return 0; | |
720 | } | |
721 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
722 | ||
723 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
724 | unsigned long len) | |
725 | { | |
726 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
727 | int seg; | |
728 | int offset = offset_in_page(gpa); | |
729 | int ret; | |
730 | ||
731 | while ((seg = next_segment(len, offset)) != 0) { | |
732 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
733 | if (ret < 0) | |
734 | return ret; | |
735 | offset = 0; | |
736 | len -= seg; | |
737 | data += seg; | |
738 | ++gfn; | |
739 | } | |
740 | return 0; | |
741 | } | |
742 | ||
743 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
744 | { | |
3e021bf5 | 745 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
746 | } |
747 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
748 | ||
749 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
750 | { | |
751 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
752 | int seg; | |
753 | int offset = offset_in_page(gpa); | |
754 | int ret; | |
755 | ||
756 | while ((seg = next_segment(len, offset)) != 0) { | |
757 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
758 | if (ret < 0) | |
759 | return ret; | |
760 | offset = 0; | |
761 | len -= seg; | |
762 | ++gfn; | |
763 | } | |
764 | return 0; | |
765 | } | |
766 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
767 | ||
6aa8b732 AK |
768 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
769 | { | |
31389947 | 770 | struct kvm_memory_slot *memslot; |
6aa8b732 | 771 | |
3b6fff19 | 772 | gfn = unalias_gfn(kvm, gfn); |
7e9d619d RR |
773 | memslot = __gfn_to_memslot(kvm, gfn); |
774 | if (memslot && memslot->dirty_bitmap) { | |
775 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 776 | |
7e9d619d RR |
777 | /* avoid RMW */ |
778 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
779 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
780 | } |
781 | } | |
782 | ||
b6958ce4 ED |
783 | /* |
784 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
785 | */ | |
8776e519 | 786 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 787 | { |
e5c239cf MT |
788 | DEFINE_WAIT(wait); |
789 | ||
790 | for (;;) { | |
791 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
792 | ||
793 | if (kvm_cpu_has_interrupt(vcpu)) | |
794 | break; | |
795 | if (kvm_cpu_has_pending_timer(vcpu)) | |
796 | break; | |
797 | if (kvm_arch_vcpu_runnable(vcpu)) | |
798 | break; | |
799 | if (signal_pending(current)) | |
800 | break; | |
801 | ||
b6958ce4 ED |
802 | vcpu_put(vcpu); |
803 | schedule(); | |
804 | vcpu_load(vcpu); | |
805 | } | |
d3bef15f | 806 | |
e5c239cf | 807 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
808 | } |
809 | ||
6aa8b732 AK |
810 | void kvm_resched(struct kvm_vcpu *vcpu) |
811 | { | |
3fca0365 YD |
812 | if (!need_resched()) |
813 | return; | |
6aa8b732 | 814 | cond_resched(); |
6aa8b732 AK |
815 | } |
816 | EXPORT_SYMBOL_GPL(kvm_resched); | |
817 | ||
e4a533a4 | 818 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
819 | { |
820 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
821 | struct page *page; |
822 | ||
e4a533a4 | 823 | if (vmf->pgoff == 0) |
039576c0 | 824 | page = virt_to_page(vcpu->run); |
09566765 | 825 | #ifdef CONFIG_X86 |
e4a533a4 | 826 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 827 | page = virt_to_page(vcpu->arch.pio_data); |
09566765 | 828 | #endif |
039576c0 | 829 | else |
e4a533a4 | 830 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 831 | get_page(page); |
e4a533a4 | 832 | vmf->page = page; |
833 | return 0; | |
9a2bb7f4 AK |
834 | } |
835 | ||
836 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
e4a533a4 | 837 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
838 | }; |
839 | ||
840 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
841 | { | |
842 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
843 | return 0; | |
844 | } | |
845 | ||
bccf2150 AK |
846 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
847 | { | |
848 | struct kvm_vcpu *vcpu = filp->private_data; | |
849 | ||
66c0b394 | 850 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
851 | return 0; |
852 | } | |
853 | ||
5c502742 | 854 | static const struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
855 | .release = kvm_vcpu_release, |
856 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
857 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 858 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
859 | }; |
860 | ||
861 | /* | |
862 | * Allocates an inode for the vcpu. | |
863 | */ | |
864 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
865 | { | |
2030a42c AV |
866 | int fd = anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu); |
867 | if (fd < 0) | |
66c0b394 | 868 | kvm_put_kvm(vcpu->kvm); |
bccf2150 | 869 | return fd; |
bccf2150 AK |
870 | } |
871 | ||
c5ea7660 AK |
872 | /* |
873 | * Creates some virtual cpus. Good luck creating more than one. | |
874 | */ | |
875 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
876 | { | |
877 | int r; | |
878 | struct kvm_vcpu *vcpu; | |
879 | ||
c5ea7660 | 880 | if (!valid_vcpu(n)) |
fb3f0f51 | 881 | return -EINVAL; |
c5ea7660 | 882 | |
e9b11c17 | 883 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
884 | if (IS_ERR(vcpu)) |
885 | return PTR_ERR(vcpu); | |
c5ea7660 | 886 | |
15ad7146 AK |
887 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
888 | ||
26e5215f AK |
889 | r = kvm_arch_vcpu_setup(vcpu); |
890 | if (r) | |
891 | goto vcpu_destroy; | |
892 | ||
11ec2804 | 893 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
894 | if (kvm->vcpus[n]) { |
895 | r = -EEXIST; | |
11ec2804 | 896 | mutex_unlock(&kvm->lock); |
e9b11c17 | 897 | goto vcpu_destroy; |
fb3f0f51 RR |
898 | } |
899 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 900 | mutex_unlock(&kvm->lock); |
c5ea7660 | 901 | |
fb3f0f51 | 902 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 903 | kvm_get_kvm(kvm); |
bccf2150 AK |
904 | r = create_vcpu_fd(vcpu); |
905 | if (r < 0) | |
fb3f0f51 RR |
906 | goto unlink; |
907 | return r; | |
39c3b86e | 908 | |
fb3f0f51 | 909 | unlink: |
11ec2804 | 910 | mutex_lock(&kvm->lock); |
fb3f0f51 | 911 | kvm->vcpus[n] = NULL; |
11ec2804 | 912 | mutex_unlock(&kvm->lock); |
e9b11c17 | 913 | vcpu_destroy: |
d40ccc62 | 914 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
915 | return r; |
916 | } | |
917 | ||
1961d276 AK |
918 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
919 | { | |
920 | if (sigset) { | |
921 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
922 | vcpu->sigset_active = 1; | |
923 | vcpu->sigset = *sigset; | |
924 | } else | |
925 | vcpu->sigset_active = 0; | |
926 | return 0; | |
927 | } | |
928 | ||
bccf2150 AK |
929 | static long kvm_vcpu_ioctl(struct file *filp, |
930 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 931 | { |
bccf2150 | 932 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 933 | void __user *argp = (void __user *)arg; |
313a3dc7 | 934 | int r; |
6aa8b732 | 935 | |
6d4e4c4f AK |
936 | if (vcpu->kvm->mm != current->mm) |
937 | return -EIO; | |
6aa8b732 | 938 | switch (ioctl) { |
9a2bb7f4 | 939 | case KVM_RUN: |
f0fe5108 AK |
940 | r = -EINVAL; |
941 | if (arg) | |
942 | goto out; | |
b6c7a5dc | 943 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 944 | break; |
6aa8b732 | 945 | case KVM_GET_REGS: { |
3e4bb3ac | 946 | struct kvm_regs *kvm_regs; |
6aa8b732 | 947 | |
3e4bb3ac XZ |
948 | r = -ENOMEM; |
949 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
950 | if (!kvm_regs) | |
6aa8b732 | 951 | goto out; |
3e4bb3ac XZ |
952 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
953 | if (r) | |
954 | goto out_free1; | |
6aa8b732 | 955 | r = -EFAULT; |
3e4bb3ac XZ |
956 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
957 | goto out_free1; | |
6aa8b732 | 958 | r = 0; |
3e4bb3ac XZ |
959 | out_free1: |
960 | kfree(kvm_regs); | |
6aa8b732 AK |
961 | break; |
962 | } | |
963 | case KVM_SET_REGS: { | |
3e4bb3ac | 964 | struct kvm_regs *kvm_regs; |
6aa8b732 | 965 | |
3e4bb3ac XZ |
966 | r = -ENOMEM; |
967 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
968 | if (!kvm_regs) | |
6aa8b732 | 969 | goto out; |
3e4bb3ac XZ |
970 | r = -EFAULT; |
971 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
972 | goto out_free2; | |
973 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 974 | if (r) |
3e4bb3ac | 975 | goto out_free2; |
6aa8b732 | 976 | r = 0; |
3e4bb3ac XZ |
977 | out_free2: |
978 | kfree(kvm_regs); | |
6aa8b732 AK |
979 | break; |
980 | } | |
981 | case KVM_GET_SREGS: { | |
982 | struct kvm_sregs kvm_sregs; | |
983 | ||
bccf2150 | 984 | memset(&kvm_sregs, 0, sizeof kvm_sregs); |
b6c7a5dc | 985 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
986 | if (r) |
987 | goto out; | |
988 | r = -EFAULT; | |
2f366987 | 989 | if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs)) |
6aa8b732 AK |
990 | goto out; |
991 | r = 0; | |
992 | break; | |
993 | } | |
994 | case KVM_SET_SREGS: { | |
995 | struct kvm_sregs kvm_sregs; | |
996 | ||
997 | r = -EFAULT; | |
2f366987 | 998 | if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs)) |
6aa8b732 | 999 | goto out; |
b6c7a5dc | 1000 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
1001 | if (r) |
1002 | goto out; | |
1003 | r = 0; | |
1004 | break; | |
1005 | } | |
62d9f0db MT |
1006 | case KVM_GET_MP_STATE: { |
1007 | struct kvm_mp_state mp_state; | |
1008 | ||
1009 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1010 | if (r) | |
1011 | goto out; | |
1012 | r = -EFAULT; | |
1013 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1014 | goto out; | |
1015 | r = 0; | |
1016 | break; | |
1017 | } | |
1018 | case KVM_SET_MP_STATE: { | |
1019 | struct kvm_mp_state mp_state; | |
1020 | ||
1021 | r = -EFAULT; | |
1022 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1023 | goto out; | |
1024 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1025 | if (r) | |
1026 | goto out; | |
1027 | r = 0; | |
1028 | break; | |
1029 | } | |
6aa8b732 AK |
1030 | case KVM_TRANSLATE: { |
1031 | struct kvm_translation tr; | |
1032 | ||
1033 | r = -EFAULT; | |
2f366987 | 1034 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1035 | goto out; |
8b006791 | 1036 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1037 | if (r) |
1038 | goto out; | |
1039 | r = -EFAULT; | |
2f366987 | 1040 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1041 | goto out; |
1042 | r = 0; | |
1043 | break; | |
1044 | } | |
6aa8b732 AK |
1045 | case KVM_DEBUG_GUEST: { |
1046 | struct kvm_debug_guest dbg; | |
1047 | ||
1048 | r = -EFAULT; | |
2f366987 | 1049 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1050 | goto out; |
b6c7a5dc | 1051 | r = kvm_arch_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
1052 | if (r) |
1053 | goto out; | |
1054 | r = 0; | |
1055 | break; | |
1056 | } | |
1961d276 AK |
1057 | case KVM_SET_SIGNAL_MASK: { |
1058 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1059 | struct kvm_signal_mask kvm_sigmask; | |
1060 | sigset_t sigset, *p; | |
1061 | ||
1062 | p = NULL; | |
1063 | if (argp) { | |
1064 | r = -EFAULT; | |
1065 | if (copy_from_user(&kvm_sigmask, argp, | |
1066 | sizeof kvm_sigmask)) | |
1067 | goto out; | |
1068 | r = -EINVAL; | |
1069 | if (kvm_sigmask.len != sizeof sigset) | |
1070 | goto out; | |
1071 | r = -EFAULT; | |
1072 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1073 | sizeof sigset)) | |
1074 | goto out; | |
1075 | p = &sigset; | |
1076 | } | |
1077 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1078 | break; | |
1079 | } | |
b8836737 AK |
1080 | case KVM_GET_FPU: { |
1081 | struct kvm_fpu fpu; | |
1082 | ||
1083 | memset(&fpu, 0, sizeof fpu); | |
d0752060 | 1084 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, &fpu); |
b8836737 AK |
1085 | if (r) |
1086 | goto out; | |
1087 | r = -EFAULT; | |
1088 | if (copy_to_user(argp, &fpu, sizeof fpu)) | |
1089 | goto out; | |
1090 | r = 0; | |
1091 | break; | |
1092 | } | |
1093 | case KVM_SET_FPU: { | |
1094 | struct kvm_fpu fpu; | |
1095 | ||
1096 | r = -EFAULT; | |
1097 | if (copy_from_user(&fpu, argp, sizeof fpu)) | |
1098 | goto out; | |
d0752060 | 1099 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, &fpu); |
b8836737 AK |
1100 | if (r) |
1101 | goto out; | |
1102 | r = 0; | |
1103 | break; | |
1104 | } | |
bccf2150 | 1105 | default: |
313a3dc7 | 1106 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1107 | } |
1108 | out: | |
1109 | return r; | |
1110 | } | |
1111 | ||
1112 | static long kvm_vm_ioctl(struct file *filp, | |
1113 | unsigned int ioctl, unsigned long arg) | |
1114 | { | |
1115 | struct kvm *kvm = filp->private_data; | |
1116 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1117 | int r; |
bccf2150 | 1118 | |
6d4e4c4f AK |
1119 | if (kvm->mm != current->mm) |
1120 | return -EIO; | |
bccf2150 AK |
1121 | switch (ioctl) { |
1122 | case KVM_CREATE_VCPU: | |
1123 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1124 | if (r < 0) | |
1125 | goto out; | |
1126 | break; | |
6fc138d2 IE |
1127 | case KVM_SET_USER_MEMORY_REGION: { |
1128 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1129 | ||
1130 | r = -EFAULT; | |
1131 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1132 | sizeof kvm_userspace_mem)) | |
1133 | goto out; | |
1134 | ||
1135 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1136 | if (r) |
1137 | goto out; | |
1138 | break; | |
1139 | } | |
1140 | case KVM_GET_DIRTY_LOG: { | |
1141 | struct kvm_dirty_log log; | |
1142 | ||
1143 | r = -EFAULT; | |
2f366987 | 1144 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1145 | goto out; |
2c6f5df9 | 1146 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1147 | if (r) |
1148 | goto out; | |
1149 | break; | |
1150 | } | |
f17abe9a | 1151 | default: |
1fe779f8 | 1152 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
1153 | } |
1154 | out: | |
1155 | return r; | |
1156 | } | |
1157 | ||
e4a533a4 | 1158 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a AK |
1159 | { |
1160 | struct kvm *kvm = vma->vm_file->private_data; | |
f17abe9a AK |
1161 | struct page *page; |
1162 | ||
e4a533a4 | 1163 | if (!kvm_is_visible_gfn(kvm, vmf->pgoff)) |
1164 | return VM_FAULT_SIGBUS; | |
10589a46 | 1165 | page = gfn_to_page(kvm, vmf->pgoff); |
8a7ae055 | 1166 | if (is_error_page(page)) { |
b4231d61 | 1167 | kvm_release_page_clean(page); |
e4a533a4 | 1168 | return VM_FAULT_SIGBUS; |
8a7ae055 | 1169 | } |
e4a533a4 | 1170 | vmf->page = page; |
1171 | return 0; | |
f17abe9a AK |
1172 | } |
1173 | ||
1174 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
e4a533a4 | 1175 | .fault = kvm_vm_fault, |
f17abe9a AK |
1176 | }; |
1177 | ||
1178 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1179 | { | |
1180 | vma->vm_ops = &kvm_vm_vm_ops; | |
1181 | return 0; | |
1182 | } | |
1183 | ||
5c502742 | 1184 | static const struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1185 | .release = kvm_vm_release, |
1186 | .unlocked_ioctl = kvm_vm_ioctl, | |
1187 | .compat_ioctl = kvm_vm_ioctl, | |
1188 | .mmap = kvm_vm_mmap, | |
1189 | }; | |
1190 | ||
1191 | static int kvm_dev_ioctl_create_vm(void) | |
1192 | { | |
2030a42c | 1193 | int fd; |
f17abe9a AK |
1194 | struct kvm *kvm; |
1195 | ||
f17abe9a | 1196 | kvm = kvm_create_vm(); |
d6d28168 AK |
1197 | if (IS_ERR(kvm)) |
1198 | return PTR_ERR(kvm); | |
2030a42c AV |
1199 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm); |
1200 | if (fd < 0) | |
66c0b394 | 1201 | kvm_put_kvm(kvm); |
f17abe9a | 1202 | |
f17abe9a | 1203 | return fd; |
f17abe9a AK |
1204 | } |
1205 | ||
1206 | static long kvm_dev_ioctl(struct file *filp, | |
1207 | unsigned int ioctl, unsigned long arg) | |
1208 | { | |
07c45a36 | 1209 | long r = -EINVAL; |
f17abe9a AK |
1210 | |
1211 | switch (ioctl) { | |
1212 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1213 | r = -EINVAL; |
1214 | if (arg) | |
1215 | goto out; | |
f17abe9a AK |
1216 | r = KVM_API_VERSION; |
1217 | break; | |
1218 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1219 | r = -EINVAL; |
1220 | if (arg) | |
1221 | goto out; | |
f17abe9a AK |
1222 | r = kvm_dev_ioctl_create_vm(); |
1223 | break; | |
018d00d2 | 1224 | case KVM_CHECK_EXTENSION: |
1e1c65e0 | 1225 | r = kvm_dev_ioctl_check_extension(arg); |
5d308f45 | 1226 | break; |
07c45a36 AK |
1227 | case KVM_GET_VCPU_MMAP_SIZE: |
1228 | r = -EINVAL; | |
1229 | if (arg) | |
1230 | goto out; | |
adb1ff46 AK |
1231 | r = PAGE_SIZE; /* struct kvm_run */ |
1232 | #ifdef CONFIG_X86 | |
1233 | r += PAGE_SIZE; /* pio data page */ | |
1234 | #endif | |
07c45a36 | 1235 | break; |
d4c9ff2d FEL |
1236 | case KVM_TRACE_ENABLE: |
1237 | case KVM_TRACE_PAUSE: | |
1238 | case KVM_TRACE_DISABLE: | |
1239 | r = kvm_trace_ioctl(ioctl, arg); | |
1240 | break; | |
6aa8b732 | 1241 | default: |
043405e1 | 1242 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1243 | } |
1244 | out: | |
1245 | return r; | |
1246 | } | |
1247 | ||
6aa8b732 | 1248 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1249 | .unlocked_ioctl = kvm_dev_ioctl, |
1250 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1251 | }; |
1252 | ||
1253 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1254 | KVM_MINOR, |
6aa8b732 AK |
1255 | "kvm", |
1256 | &kvm_chardev_ops, | |
1257 | }; | |
1258 | ||
1b6c0168 AK |
1259 | static void hardware_enable(void *junk) |
1260 | { | |
1261 | int cpu = raw_smp_processor_id(); | |
1262 | ||
1263 | if (cpu_isset(cpu, cpus_hardware_enabled)) | |
1264 | return; | |
1265 | cpu_set(cpu, cpus_hardware_enabled); | |
e9b11c17 | 1266 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
1267 | } |
1268 | ||
1269 | static void hardware_disable(void *junk) | |
1270 | { | |
1271 | int cpu = raw_smp_processor_id(); | |
1272 | ||
1273 | if (!cpu_isset(cpu, cpus_hardware_enabled)) | |
1274 | return; | |
1275 | cpu_clear(cpu, cpus_hardware_enabled); | |
e9b11c17 | 1276 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1277 | } |
1278 | ||
774c47f1 AK |
1279 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1280 | void *v) | |
1281 | { | |
1282 | int cpu = (long)v; | |
1283 | ||
1a6f4d7f | 1284 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1285 | switch (val) { |
cec9ad27 | 1286 | case CPU_DYING: |
6ec8a856 AK |
1287 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1288 | cpu); | |
1289 | hardware_disable(NULL); | |
1290 | break; | |
774c47f1 | 1291 | case CPU_UP_CANCELED: |
43934a38 JK |
1292 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1293 | cpu); | |
8691e5a8 | 1294 | smp_call_function_single(cpu, hardware_disable, NULL, 1); |
774c47f1 | 1295 | break; |
43934a38 JK |
1296 | case CPU_ONLINE: |
1297 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1298 | cpu); | |
8691e5a8 | 1299 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
1300 | break; |
1301 | } | |
1302 | return NOTIFY_OK; | |
1303 | } | |
1304 | ||
4ecac3fd AK |
1305 | |
1306 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
1307 | { | |
1308 | if (kvm_rebooting) | |
1309 | /* spin while reset goes on */ | |
1310 | while (true) | |
1311 | ; | |
1312 | /* Fault while not rebooting. We want the trace. */ | |
1313 | BUG(); | |
1314 | } | |
1315 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
1316 | ||
9a2b85c6 | 1317 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1318 | void *v) |
9a2b85c6 RR |
1319 | { |
1320 | if (val == SYS_RESTART) { | |
1321 | /* | |
1322 | * Some (well, at least mine) BIOSes hang on reboot if | |
1323 | * in vmx root mode. | |
1324 | */ | |
1325 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
4ecac3fd | 1326 | kvm_rebooting = true; |
15c8b6c1 | 1327 | on_each_cpu(hardware_disable, NULL, 1); |
9a2b85c6 RR |
1328 | } |
1329 | return NOTIFY_OK; | |
1330 | } | |
1331 | ||
1332 | static struct notifier_block kvm_reboot_notifier = { | |
1333 | .notifier_call = kvm_reboot, | |
1334 | .priority = 0, | |
1335 | }; | |
1336 | ||
2eeb2e94 GH |
1337 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
1338 | { | |
1339 | memset(bus, 0, sizeof(*bus)); | |
1340 | } | |
1341 | ||
1342 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
1343 | { | |
1344 | int i; | |
1345 | ||
1346 | for (i = 0; i < bus->dev_count; i++) { | |
1347 | struct kvm_io_device *pos = bus->devs[i]; | |
1348 | ||
1349 | kvm_iodevice_destructor(pos); | |
1350 | } | |
1351 | } | |
1352 | ||
92760499 LV |
1353 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, |
1354 | gpa_t addr, int len, int is_write) | |
2eeb2e94 GH |
1355 | { |
1356 | int i; | |
1357 | ||
1358 | for (i = 0; i < bus->dev_count; i++) { | |
1359 | struct kvm_io_device *pos = bus->devs[i]; | |
1360 | ||
92760499 | 1361 | if (pos->in_range(pos, addr, len, is_write)) |
2eeb2e94 GH |
1362 | return pos; |
1363 | } | |
1364 | ||
1365 | return NULL; | |
1366 | } | |
1367 | ||
1368 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
1369 | { | |
1370 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
1371 | ||
1372 | bus->devs[bus->dev_count++] = dev; | |
1373 | } | |
1374 | ||
774c47f1 AK |
1375 | static struct notifier_block kvm_cpu_notifier = { |
1376 | .notifier_call = kvm_cpu_hotplug, | |
1377 | .priority = 20, /* must be > scheduler priority */ | |
1378 | }; | |
1379 | ||
8b88b099 | 1380 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
1381 | { |
1382 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
1383 | struct kvm *kvm; |
1384 | ||
8b88b099 | 1385 | *val = 0; |
ba1389b7 AK |
1386 | spin_lock(&kvm_lock); |
1387 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 1388 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 1389 | spin_unlock(&kvm_lock); |
8b88b099 | 1390 | return 0; |
ba1389b7 AK |
1391 | } |
1392 | ||
1393 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
1394 | ||
8b88b099 | 1395 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
1396 | { |
1397 | unsigned offset = (long)_offset; | |
1165f5fe AK |
1398 | struct kvm *kvm; |
1399 | struct kvm_vcpu *vcpu; | |
1400 | int i; | |
1401 | ||
8b88b099 | 1402 | *val = 0; |
1165f5fe AK |
1403 | spin_lock(&kvm_lock); |
1404 | list_for_each_entry(kvm, &vm_list, vm_list) | |
1405 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
1406 | vcpu = kvm->vcpus[i]; |
1407 | if (vcpu) | |
8b88b099 | 1408 | *val += *(u32 *)((void *)vcpu + offset); |
1165f5fe AK |
1409 | } |
1410 | spin_unlock(&kvm_lock); | |
8b88b099 | 1411 | return 0; |
1165f5fe AK |
1412 | } |
1413 | ||
ba1389b7 AK |
1414 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
1415 | ||
1416 | static struct file_operations *stat_fops[] = { | |
1417 | [KVM_STAT_VCPU] = &vcpu_stat_fops, | |
1418 | [KVM_STAT_VM] = &vm_stat_fops, | |
1419 | }; | |
1165f5fe | 1420 | |
a16b043c | 1421 | static void kvm_init_debug(void) |
6aa8b732 AK |
1422 | { |
1423 | struct kvm_stats_debugfs_item *p; | |
1424 | ||
76f7c879 | 1425 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 1426 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 1427 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 1428 | (void *)(long)p->offset, |
ba1389b7 | 1429 | stat_fops[p->kind]); |
6aa8b732 AK |
1430 | } |
1431 | ||
1432 | static void kvm_exit_debug(void) | |
1433 | { | |
1434 | struct kvm_stats_debugfs_item *p; | |
1435 | ||
1436 | for (p = debugfs_entries; p->name; ++p) | |
1437 | debugfs_remove(p->dentry); | |
76f7c879 | 1438 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
1439 | } |
1440 | ||
59ae6c6b AK |
1441 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
1442 | { | |
4267c41a | 1443 | hardware_disable(NULL); |
59ae6c6b AK |
1444 | return 0; |
1445 | } | |
1446 | ||
1447 | static int kvm_resume(struct sys_device *dev) | |
1448 | { | |
4267c41a | 1449 | hardware_enable(NULL); |
59ae6c6b AK |
1450 | return 0; |
1451 | } | |
1452 | ||
1453 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 1454 | .name = "kvm", |
59ae6c6b AK |
1455 | .suspend = kvm_suspend, |
1456 | .resume = kvm_resume, | |
1457 | }; | |
1458 | ||
1459 | static struct sys_device kvm_sysdev = { | |
1460 | .id = 0, | |
1461 | .cls = &kvm_sysdev_class, | |
1462 | }; | |
1463 | ||
cea7bb21 | 1464 | struct page *bad_page; |
35149e21 | 1465 | pfn_t bad_pfn; |
6aa8b732 | 1466 | |
15ad7146 AK |
1467 | static inline |
1468 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
1469 | { | |
1470 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
1471 | } | |
1472 | ||
1473 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
1474 | { | |
1475 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1476 | ||
e9b11c17 | 1477 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
1478 | } |
1479 | ||
1480 | static void kvm_sched_out(struct preempt_notifier *pn, | |
1481 | struct task_struct *next) | |
1482 | { | |
1483 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1484 | ||
e9b11c17 | 1485 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
1486 | } |
1487 | ||
f8c16bba | 1488 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 1489 | struct module *module) |
6aa8b732 AK |
1490 | { |
1491 | int r; | |
002c7f7c | 1492 | int cpu; |
6aa8b732 | 1493 | |
cb498ea2 ZX |
1494 | kvm_init_debug(); |
1495 | ||
f8c16bba ZX |
1496 | r = kvm_arch_init(opaque); |
1497 | if (r) | |
d2308784 | 1498 | goto out_fail; |
cb498ea2 ZX |
1499 | |
1500 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
1501 | ||
1502 | if (bad_page == NULL) { | |
1503 | r = -ENOMEM; | |
1504 | goto out; | |
1505 | } | |
1506 | ||
35149e21 AL |
1507 | bad_pfn = page_to_pfn(bad_page); |
1508 | ||
e9b11c17 | 1509 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 1510 | if (r < 0) |
d2308784 | 1511 | goto out_free_0; |
6aa8b732 | 1512 | |
002c7f7c YS |
1513 | for_each_online_cpu(cpu) { |
1514 | smp_call_function_single(cpu, | |
e9b11c17 | 1515 | kvm_arch_check_processor_compat, |
8691e5a8 | 1516 | &r, 1); |
002c7f7c | 1517 | if (r < 0) |
d2308784 | 1518 | goto out_free_1; |
002c7f7c YS |
1519 | } |
1520 | ||
15c8b6c1 | 1521 | on_each_cpu(hardware_enable, NULL, 1); |
774c47f1 AK |
1522 | r = register_cpu_notifier(&kvm_cpu_notifier); |
1523 | if (r) | |
d2308784 | 1524 | goto out_free_2; |
6aa8b732 AK |
1525 | register_reboot_notifier(&kvm_reboot_notifier); |
1526 | ||
59ae6c6b AK |
1527 | r = sysdev_class_register(&kvm_sysdev_class); |
1528 | if (r) | |
d2308784 | 1529 | goto out_free_3; |
59ae6c6b AK |
1530 | |
1531 | r = sysdev_register(&kvm_sysdev); | |
1532 | if (r) | |
d2308784 | 1533 | goto out_free_4; |
59ae6c6b | 1534 | |
c16f862d RR |
1535 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
1536 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
1537 | __alignof__(struct kvm_vcpu), |
1538 | 0, NULL); | |
c16f862d RR |
1539 | if (!kvm_vcpu_cache) { |
1540 | r = -ENOMEM; | |
d2308784 | 1541 | goto out_free_5; |
c16f862d RR |
1542 | } |
1543 | ||
6aa8b732 AK |
1544 | kvm_chardev_ops.owner = module; |
1545 | ||
1546 | r = misc_register(&kvm_dev); | |
1547 | if (r) { | |
d77c26fc | 1548 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
1549 | goto out_free; |
1550 | } | |
1551 | ||
15ad7146 AK |
1552 | kvm_preempt_ops.sched_in = kvm_sched_in; |
1553 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
1554 | ||
c7addb90 | 1555 | return 0; |
6aa8b732 AK |
1556 | |
1557 | out_free: | |
c16f862d | 1558 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 1559 | out_free_5: |
59ae6c6b | 1560 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 1561 | out_free_4: |
59ae6c6b | 1562 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 1563 | out_free_3: |
6aa8b732 | 1564 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 1565 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 1566 | out_free_2: |
15c8b6c1 | 1567 | on_each_cpu(hardware_disable, NULL, 1); |
d2308784 | 1568 | out_free_1: |
e9b11c17 | 1569 | kvm_arch_hardware_unsetup(); |
d2308784 ZX |
1570 | out_free_0: |
1571 | __free_page(bad_page); | |
ca45aaae | 1572 | out: |
f8c16bba | 1573 | kvm_arch_exit(); |
cb498ea2 | 1574 | kvm_exit_debug(); |
d2308784 | 1575 | out_fail: |
6aa8b732 AK |
1576 | return r; |
1577 | } | |
cb498ea2 | 1578 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 1579 | |
cb498ea2 | 1580 | void kvm_exit(void) |
6aa8b732 | 1581 | { |
d4c9ff2d | 1582 | kvm_trace_cleanup(); |
6aa8b732 | 1583 | misc_deregister(&kvm_dev); |
c16f862d | 1584 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
1585 | sysdev_unregister(&kvm_sysdev); |
1586 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 1587 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 1588 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 1589 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 1590 | kvm_arch_hardware_unsetup(); |
f8c16bba | 1591 | kvm_arch_exit(); |
6aa8b732 | 1592 | kvm_exit_debug(); |
cea7bb21 | 1593 | __free_page(bad_page); |
6aa8b732 | 1594 | } |
cb498ea2 | 1595 | EXPORT_SYMBOL_GPL(kvm_exit); |