]>
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 | ||
18 | #include "kvm.h" | |
043405e1 | 19 | #include "x86.h" |
85f455f7 | 20 | #include "irq.h" |
6aa8b732 AK |
21 | |
22 | #include <linux/kvm.h> | |
23 | #include <linux/module.h> | |
24 | #include <linux/errno.h> | |
6aa8b732 AK |
25 | #include <linux/percpu.h> |
26 | #include <linux/gfp.h> | |
6aa8b732 AK |
27 | #include <linux/mm.h> |
28 | #include <linux/miscdevice.h> | |
29 | #include <linux/vmalloc.h> | |
6aa8b732 | 30 | #include <linux/reboot.h> |
6aa8b732 AK |
31 | #include <linux/debugfs.h> |
32 | #include <linux/highmem.h> | |
33 | #include <linux/file.h> | |
59ae6c6b | 34 | #include <linux/sysdev.h> |
774c47f1 | 35 | #include <linux/cpu.h> |
e8edc6e0 | 36 | #include <linux/sched.h> |
d9e368d6 AK |
37 | #include <linux/cpumask.h> |
38 | #include <linux/smp.h> | |
d6d28168 | 39 | #include <linux/anon_inodes.h> |
04d2cc77 | 40 | #include <linux/profile.h> |
7aa81cc0 | 41 | #include <linux/kvm_para.h> |
6fc138d2 | 42 | #include <linux/pagemap.h> |
8d4e1288 | 43 | #include <linux/mman.h> |
6aa8b732 | 44 | |
e495606d | 45 | #include <asm/processor.h> |
e495606d AK |
46 | #include <asm/io.h> |
47 | #include <asm/uaccess.h> | |
48 | #include <asm/desc.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 | ||
6aa8b732 AK |
63 | static struct dentry *debugfs_dir; |
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 | } |
49d3bd7e | 117 | smp_call_function_mask(cpus, ack_flush, NULL, 1); |
d9e368d6 AK |
118 | } |
119 | ||
fb3f0f51 RR |
120 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
121 | { | |
122 | struct page *page; | |
123 | int r; | |
124 | ||
125 | mutex_init(&vcpu->mutex); | |
126 | vcpu->cpu = -1; | |
fb3f0f51 RR |
127 | vcpu->kvm = kvm; |
128 | vcpu->vcpu_id = id; | |
b6958ce4 | 129 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
130 | |
131 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
132 | if (!page) { | |
133 | r = -ENOMEM; | |
134 | goto fail; | |
135 | } | |
136 | vcpu->run = page_address(page); | |
137 | ||
e9b11c17 | 138 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 139 | if (r < 0) |
e9b11c17 | 140 | goto fail_free_run; |
fb3f0f51 RR |
141 | return 0; |
142 | ||
fb3f0f51 RR |
143 | fail_free_run: |
144 | free_page((unsigned long)vcpu->run); | |
145 | fail: | |
76fafa5e | 146 | return r; |
fb3f0f51 RR |
147 | } |
148 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
149 | ||
150 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
151 | { | |
e9b11c17 | 152 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
153 | free_page((unsigned long)vcpu->run); |
154 | } | |
155 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
156 | ||
f17abe9a | 157 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 158 | { |
d19a9cd2 | 159 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 160 | |
d19a9cd2 ZX |
161 | if (IS_ERR(kvm)) |
162 | goto out; | |
6aa8b732 | 163 | |
74906345 | 164 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 165 | mutex_init(&kvm->lock); |
2eeb2e94 | 166 | kvm_io_bus_init(&kvm->mmio_bus); |
5e58cfe4 RR |
167 | spin_lock(&kvm_lock); |
168 | list_add(&kvm->vm_list, &vm_list); | |
169 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 170 | out: |
f17abe9a AK |
171 | return kvm; |
172 | } | |
173 | ||
6aa8b732 AK |
174 | /* |
175 | * Free any memory in @free but not in @dont. | |
176 | */ | |
177 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
178 | struct kvm_memory_slot *dont) | |
179 | { | |
290fc38d IE |
180 | if (!dont || free->rmap != dont->rmap) |
181 | vfree(free->rmap); | |
6aa8b732 AK |
182 | |
183 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
184 | vfree(free->dirty_bitmap); | |
185 | ||
6aa8b732 | 186 | free->npages = 0; |
8b6d44c7 | 187 | free->dirty_bitmap = NULL; |
8d4e1288 | 188 | free->rmap = NULL; |
6aa8b732 AK |
189 | } |
190 | ||
d19a9cd2 | 191 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
192 | { |
193 | int i; | |
194 | ||
195 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 196 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
197 | } |
198 | ||
f17abe9a AK |
199 | static void kvm_destroy_vm(struct kvm *kvm) |
200 | { | |
133de902 AK |
201 | spin_lock(&kvm_lock); |
202 | list_del(&kvm->vm_list); | |
203 | spin_unlock(&kvm_lock); | |
74906345 | 204 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 205 | kvm_io_bus_destroy(&kvm->mmio_bus); |
d19a9cd2 | 206 | kvm_arch_destroy_vm(kvm); |
f17abe9a AK |
207 | } |
208 | ||
209 | static int kvm_vm_release(struct inode *inode, struct file *filp) | |
210 | { | |
211 | struct kvm *kvm = filp->private_data; | |
212 | ||
213 | kvm_destroy_vm(kvm); | |
6aa8b732 AK |
214 | return 0; |
215 | } | |
216 | ||
6aa8b732 AK |
217 | /* |
218 | * Allocate some memory and give it an address in the guest physical address | |
219 | * space. | |
220 | * | |
221 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e SY |
222 | * |
223 | * Must be called holding kvm->lock. | |
6aa8b732 | 224 | */ |
f78e0e2e SY |
225 | int __kvm_set_memory_region(struct kvm *kvm, |
226 | struct kvm_userspace_memory_region *mem, | |
227 | int user_alloc) | |
6aa8b732 AK |
228 | { |
229 | int r; | |
230 | gfn_t base_gfn; | |
231 | unsigned long npages; | |
232 | unsigned long i; | |
233 | struct kvm_memory_slot *memslot; | |
234 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
235 | |
236 | r = -EINVAL; | |
237 | /* General sanity checks */ | |
238 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
239 | goto out; | |
240 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
241 | goto out; | |
e0d62c7f | 242 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
243 | goto out; |
244 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
245 | goto out; | |
246 | ||
247 | memslot = &kvm->memslots[mem->slot]; | |
248 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
249 | npages = mem->memory_size >> PAGE_SHIFT; | |
250 | ||
251 | if (!npages) | |
252 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
253 | ||
6aa8b732 AK |
254 | new = old = *memslot; |
255 | ||
256 | new.base_gfn = base_gfn; | |
257 | new.npages = npages; | |
258 | new.flags = mem->flags; | |
259 | ||
260 | /* Disallow changing a memory slot's size. */ | |
261 | r = -EINVAL; | |
262 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 263 | goto out_free; |
6aa8b732 AK |
264 | |
265 | /* Check for overlaps */ | |
266 | r = -EEXIST; | |
267 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
268 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
269 | ||
270 | if (s == memslot) | |
271 | continue; | |
272 | if (!((base_gfn + npages <= s->base_gfn) || | |
273 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 274 | goto out_free; |
6aa8b732 | 275 | } |
6aa8b732 | 276 | |
6aa8b732 AK |
277 | /* Free page dirty bitmap if unneeded */ |
278 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 279 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
280 | |
281 | r = -ENOMEM; | |
282 | ||
283 | /* Allocate if a slot is being created */ | |
8d4e1288 | 284 | if (npages && !new.rmap) { |
d77c26fc | 285 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
286 | |
287 | if (!new.rmap) | |
f78e0e2e | 288 | goto out_free; |
290fc38d | 289 | |
290fc38d | 290 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 291 | |
80b14b5b | 292 | new.user_alloc = user_alloc; |
8d4e1288 | 293 | if (user_alloc) |
8a7ae055 | 294 | new.userspace_addr = mem->userspace_addr; |
8d4e1288 AL |
295 | else { |
296 | down_write(¤t->mm->mmap_sem); | |
297 | new.userspace_addr = do_mmap(NULL, 0, | |
298 | npages * PAGE_SIZE, | |
299 | PROT_READ | PROT_WRITE, | |
300 | MAP_SHARED | MAP_ANONYMOUS, | |
301 | 0); | |
302 | up_write(¤t->mm->mmap_sem); | |
303 | ||
304 | if (IS_ERR((void *)new.userspace_addr)) | |
f78e0e2e | 305 | goto out_free; |
6aa8b732 | 306 | } |
80b14b5b IE |
307 | } else { |
308 | if (!old.user_alloc && old.rmap) { | |
309 | int ret; | |
310 | ||
311 | down_write(¤t->mm->mmap_sem); | |
312 | ret = do_munmap(current->mm, old.userspace_addr, | |
313 | old.npages * PAGE_SIZE); | |
314 | up_write(¤t->mm->mmap_sem); | |
315 | if (ret < 0) | |
316 | printk(KERN_WARNING | |
317 | "kvm_vm_ioctl_set_memory_region: " | |
318 | "failed to munmap memory\n"); | |
319 | } | |
6aa8b732 AK |
320 | } |
321 | ||
322 | /* Allocate page dirty bitmap if needed */ | |
323 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
324 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
325 | ||
326 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
327 | if (!new.dirty_bitmap) | |
f78e0e2e | 328 | goto out_free; |
6aa8b732 AK |
329 | memset(new.dirty_bitmap, 0, dirty_bytes); |
330 | } | |
331 | ||
6aa8b732 AK |
332 | if (mem->slot >= kvm->nmemslots) |
333 | kvm->nmemslots = mem->slot + 1; | |
334 | ||
82ce2c96 IE |
335 | if (!kvm->n_requested_mmu_pages) { |
336 | unsigned int n_pages; | |
337 | ||
338 | if (npages) { | |
339 | n_pages = npages * KVM_PERMILLE_MMU_PAGES / 1000; | |
340 | kvm_mmu_change_mmu_pages(kvm, kvm->n_alloc_mmu_pages + | |
341 | n_pages); | |
342 | } else { | |
343 | unsigned int nr_mmu_pages; | |
344 | ||
345 | n_pages = old.npages * KVM_PERMILLE_MMU_PAGES / 1000; | |
346 | nr_mmu_pages = kvm->n_alloc_mmu_pages - n_pages; | |
347 | nr_mmu_pages = max(nr_mmu_pages, | |
348 | (unsigned int) KVM_MIN_ALLOC_MMU_PAGES); | |
349 | kvm_mmu_change_mmu_pages(kvm, nr_mmu_pages); | |
350 | } | |
351 | } | |
352 | ||
6aa8b732 | 353 | *memslot = new; |
6aa8b732 | 354 | |
90cb0529 AK |
355 | kvm_mmu_slot_remove_write_access(kvm, mem->slot); |
356 | kvm_flush_remote_tlbs(kvm); | |
6aa8b732 | 357 | |
6aa8b732 AK |
358 | kvm_free_physmem_slot(&old, &new); |
359 | return 0; | |
360 | ||
f78e0e2e | 361 | out_free: |
6aa8b732 AK |
362 | kvm_free_physmem_slot(&new, &old); |
363 | out: | |
364 | return r; | |
210c7c4d IE |
365 | |
366 | } | |
f78e0e2e SY |
367 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
368 | ||
369 | int kvm_set_memory_region(struct kvm *kvm, | |
370 | struct kvm_userspace_memory_region *mem, | |
371 | int user_alloc) | |
372 | { | |
373 | int r; | |
374 | ||
375 | mutex_lock(&kvm->lock); | |
376 | r = __kvm_set_memory_region(kvm, mem, user_alloc); | |
377 | mutex_unlock(&kvm->lock); | |
378 | return r; | |
379 | } | |
210c7c4d IE |
380 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
381 | ||
1fe779f8 CO |
382 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
383 | struct | |
384 | kvm_userspace_memory_region *mem, | |
385 | int user_alloc) | |
210c7c4d | 386 | { |
e0d62c7f IE |
387 | if (mem->slot >= KVM_MEMORY_SLOTS) |
388 | return -EINVAL; | |
210c7c4d | 389 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
390 | } |
391 | ||
392 | /* | |
393 | * Get (and clear) the dirty memory log for a memory slot. | |
394 | */ | |
2c6f5df9 AK |
395 | static int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, |
396 | struct kvm_dirty_log *log) | |
6aa8b732 AK |
397 | { |
398 | struct kvm_memory_slot *memslot; | |
399 | int r, i; | |
400 | int n; | |
401 | unsigned long any = 0; | |
402 | ||
11ec2804 | 403 | mutex_lock(&kvm->lock); |
6aa8b732 | 404 | |
6aa8b732 AK |
405 | r = -EINVAL; |
406 | if (log->slot >= KVM_MEMORY_SLOTS) | |
407 | goto out; | |
408 | ||
409 | memslot = &kvm->memslots[log->slot]; | |
410 | r = -ENOENT; | |
411 | if (!memslot->dirty_bitmap) | |
412 | goto out; | |
413 | ||
cd1a4a98 | 414 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 415 | |
cd1a4a98 | 416 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
417 | any = memslot->dirty_bitmap[i]; |
418 | ||
419 | r = -EFAULT; | |
420 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
421 | goto out; | |
422 | ||
39214915 RR |
423 | /* If nothing is dirty, don't bother messing with page tables. */ |
424 | if (any) { | |
39214915 RR |
425 | kvm_mmu_slot_remove_write_access(kvm, log->slot); |
426 | kvm_flush_remote_tlbs(kvm); | |
427 | memset(memslot->dirty_bitmap, 0, n); | |
39214915 | 428 | } |
6aa8b732 AK |
429 | |
430 | r = 0; | |
431 | ||
432 | out: | |
11ec2804 | 433 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
434 | return r; |
435 | } | |
436 | ||
cea7bb21 IE |
437 | int is_error_page(struct page *page) |
438 | { | |
439 | return page == bad_page; | |
440 | } | |
441 | EXPORT_SYMBOL_GPL(is_error_page); | |
442 | ||
f9d46eb0 IE |
443 | static inline unsigned long bad_hva(void) |
444 | { | |
445 | return PAGE_OFFSET; | |
446 | } | |
447 | ||
448 | int kvm_is_error_hva(unsigned long addr) | |
449 | { | |
450 | return addr == bad_hva(); | |
451 | } | |
452 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
453 | ||
290fc38d | 454 | gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn) |
e8207547 AK |
455 | { |
456 | int i; | |
457 | struct kvm_mem_alias *alias; | |
458 | ||
459 | for (i = 0; i < kvm->naliases; ++i) { | |
460 | alias = &kvm->aliases[i]; | |
461 | if (gfn >= alias->base_gfn | |
462 | && gfn < alias->base_gfn + alias->npages) | |
463 | return alias->target_gfn + gfn - alias->base_gfn; | |
464 | } | |
465 | return gfn; | |
466 | } | |
467 | ||
468 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
6aa8b732 AK |
469 | { |
470 | int i; | |
471 | ||
472 | for (i = 0; i < kvm->nmemslots; ++i) { | |
473 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
474 | ||
475 | if (gfn >= memslot->base_gfn | |
476 | && gfn < memslot->base_gfn + memslot->npages) | |
477 | return memslot; | |
478 | } | |
8b6d44c7 | 479 | return NULL; |
6aa8b732 | 480 | } |
e8207547 AK |
481 | |
482 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
483 | { | |
484 | gfn = unalias_gfn(kvm, gfn); | |
485 | return __gfn_to_memslot(kvm, gfn); | |
486 | } | |
6aa8b732 | 487 | |
e0d62c7f IE |
488 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
489 | { | |
490 | int i; | |
491 | ||
492 | gfn = unalias_gfn(kvm, gfn); | |
493 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
494 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
495 | ||
496 | if (gfn >= memslot->base_gfn | |
497 | && gfn < memslot->base_gfn + memslot->npages) | |
498 | return 1; | |
499 | } | |
500 | return 0; | |
501 | } | |
502 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
503 | ||
539cb660 IE |
504 | static unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
505 | { | |
506 | struct kvm_memory_slot *slot; | |
507 | ||
508 | gfn = unalias_gfn(kvm, gfn); | |
509 | slot = __gfn_to_memslot(kvm, gfn); | |
510 | if (!slot) | |
511 | return bad_hva(); | |
512 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
513 | } | |
514 | ||
aab61cc0 AL |
515 | /* |
516 | * Requires current->mm->mmap_sem to be held | |
517 | */ | |
518 | static struct page *__gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
954bbbc2 | 519 | { |
8d4e1288 | 520 | struct page *page[1]; |
539cb660 | 521 | unsigned long addr; |
8d4e1288 | 522 | int npages; |
954bbbc2 | 523 | |
60395224 AK |
524 | might_sleep(); |
525 | ||
539cb660 IE |
526 | addr = gfn_to_hva(kvm, gfn); |
527 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 528 | get_page(bad_page); |
cea7bb21 | 529 | return bad_page; |
8a7ae055 | 530 | } |
8d4e1288 | 531 | |
539cb660 IE |
532 | npages = get_user_pages(current, current->mm, addr, 1, 1, 1, page, |
533 | NULL); | |
534 | ||
8d4e1288 AL |
535 | if (npages != 1) { |
536 | get_page(bad_page); | |
537 | return bad_page; | |
8a7ae055 | 538 | } |
8d4e1288 AL |
539 | |
540 | return page[0]; | |
954bbbc2 | 541 | } |
aab61cc0 AL |
542 | |
543 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) | |
544 | { | |
545 | struct page *page; | |
546 | ||
547 | down_read(¤t->mm->mmap_sem); | |
548 | page = __gfn_to_page(kvm, gfn); | |
549 | up_read(¤t->mm->mmap_sem); | |
550 | ||
551 | return page; | |
552 | } | |
553 | ||
954bbbc2 AK |
554 | EXPORT_SYMBOL_GPL(gfn_to_page); |
555 | ||
8a7ae055 IE |
556 | void kvm_release_page(struct page *page) |
557 | { | |
558 | if (!PageReserved(page)) | |
559 | SetPageDirty(page); | |
560 | put_page(page); | |
561 | } | |
562 | EXPORT_SYMBOL_GPL(kvm_release_page); | |
563 | ||
195aefde IE |
564 | static int next_segment(unsigned long len, int offset) |
565 | { | |
566 | if (len > PAGE_SIZE - offset) | |
567 | return PAGE_SIZE - offset; | |
568 | else | |
569 | return len; | |
570 | } | |
571 | ||
572 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
573 | int len) | |
574 | { | |
e0506bcb IE |
575 | int r; |
576 | unsigned long addr; | |
195aefde | 577 | |
e0506bcb IE |
578 | addr = gfn_to_hva(kvm, gfn); |
579 | if (kvm_is_error_hva(addr)) | |
580 | return -EFAULT; | |
581 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
582 | if (r) | |
195aefde | 583 | return -EFAULT; |
195aefde IE |
584 | return 0; |
585 | } | |
586 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
587 | ||
588 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
589 | { | |
590 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
591 | int seg; | |
592 | int offset = offset_in_page(gpa); | |
593 | int ret; | |
594 | ||
595 | while ((seg = next_segment(len, offset)) != 0) { | |
596 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
597 | if (ret < 0) | |
598 | return ret; | |
599 | offset = 0; | |
600 | len -= seg; | |
601 | data += seg; | |
602 | ++gfn; | |
603 | } | |
604 | return 0; | |
605 | } | |
606 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
607 | ||
608 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, | |
609 | int offset, int len) | |
610 | { | |
e0506bcb IE |
611 | int r; |
612 | unsigned long addr; | |
195aefde | 613 | |
e0506bcb IE |
614 | addr = gfn_to_hva(kvm, gfn); |
615 | if (kvm_is_error_hva(addr)) | |
616 | return -EFAULT; | |
617 | r = copy_to_user((void __user *)addr + offset, data, len); | |
618 | if (r) | |
195aefde | 619 | return -EFAULT; |
195aefde IE |
620 | mark_page_dirty(kvm, gfn); |
621 | return 0; | |
622 | } | |
623 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
624 | ||
625 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
626 | unsigned long len) | |
627 | { | |
628 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
629 | int seg; | |
630 | int offset = offset_in_page(gpa); | |
631 | int ret; | |
632 | ||
633 | while ((seg = next_segment(len, offset)) != 0) { | |
634 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
635 | if (ret < 0) | |
636 | return ret; | |
637 | offset = 0; | |
638 | len -= seg; | |
639 | data += seg; | |
640 | ++gfn; | |
641 | } | |
642 | return 0; | |
643 | } | |
644 | ||
645 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
646 | { | |
647 | void *page_virt; | |
648 | struct page *page; | |
649 | ||
650 | page = gfn_to_page(kvm, gfn); | |
8a7ae055 IE |
651 | if (is_error_page(page)) { |
652 | kvm_release_page(page); | |
195aefde | 653 | return -EFAULT; |
8a7ae055 | 654 | } |
195aefde IE |
655 | page_virt = kmap_atomic(page, KM_USER0); |
656 | ||
657 | memset(page_virt + offset, 0, len); | |
658 | ||
659 | kunmap_atomic(page_virt, KM_USER0); | |
8a7ae055 | 660 | kvm_release_page(page); |
12264760 | 661 | mark_page_dirty(kvm, gfn); |
195aefde IE |
662 | return 0; |
663 | } | |
664 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
665 | ||
666 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, 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_clear_guest_page(kvm, gfn, offset, seg); | |
675 | if (ret < 0) | |
676 | return ret; | |
677 | offset = 0; | |
678 | len -= seg; | |
679 | ++gfn; | |
680 | } | |
681 | return 0; | |
682 | } | |
683 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
684 | ||
6aa8b732 AK |
685 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
686 | { | |
31389947 | 687 | struct kvm_memory_slot *memslot; |
6aa8b732 | 688 | |
3b6fff19 | 689 | gfn = unalias_gfn(kvm, gfn); |
7e9d619d RR |
690 | memslot = __gfn_to_memslot(kvm, gfn); |
691 | if (memslot && memslot->dirty_bitmap) { | |
692 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 693 | |
7e9d619d RR |
694 | /* avoid RMW */ |
695 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
696 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
697 | } |
698 | } | |
699 | ||
b6958ce4 ED |
700 | /* |
701 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
702 | */ | |
8776e519 | 703 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 704 | { |
b6958ce4 ED |
705 | DECLARE_WAITQUEUE(wait, current); |
706 | ||
707 | add_wait_queue(&vcpu->wq, &wait); | |
708 | ||
709 | /* | |
710 | * We will block until either an interrupt or a signal wakes us up | |
711 | */ | |
c5ec1534 HQ |
712 | while (!kvm_cpu_has_interrupt(vcpu) |
713 | && !signal_pending(current) | |
714 | && vcpu->mp_state != VCPU_MP_STATE_RUNNABLE | |
715 | && vcpu->mp_state != VCPU_MP_STATE_SIPI_RECEIVED) { | |
b6958ce4 ED |
716 | set_current_state(TASK_INTERRUPTIBLE); |
717 | vcpu_put(vcpu); | |
718 | schedule(); | |
719 | vcpu_load(vcpu); | |
720 | } | |
d3bef15f | 721 | |
c5ec1534 | 722 | __set_current_state(TASK_RUNNING); |
b6958ce4 | 723 | remove_wait_queue(&vcpu->wq, &wait); |
b6958ce4 ED |
724 | } |
725 | ||
6aa8b732 AK |
726 | void kvm_resched(struct kvm_vcpu *vcpu) |
727 | { | |
3fca0365 YD |
728 | if (!need_resched()) |
729 | return; | |
6aa8b732 | 730 | cond_resched(); |
6aa8b732 AK |
731 | } |
732 | EXPORT_SYMBOL_GPL(kvm_resched); | |
733 | ||
bccf2150 AK |
734 | static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, |
735 | struct kvm_interrupt *irq) | |
6aa8b732 | 736 | { |
6aa8b732 AK |
737 | if (irq->irq < 0 || irq->irq >= 256) |
738 | return -EINVAL; | |
97222cc8 ED |
739 | if (irqchip_in_kernel(vcpu->kvm)) |
740 | return -ENXIO; | |
bccf2150 | 741 | vcpu_load(vcpu); |
6aa8b732 AK |
742 | |
743 | set_bit(irq->irq, vcpu->irq_pending); | |
744 | set_bit(irq->irq / BITS_PER_LONG, &vcpu->irq_summary); | |
745 | ||
746 | vcpu_put(vcpu); | |
747 | ||
748 | return 0; | |
749 | } | |
750 | ||
9a2bb7f4 AK |
751 | static struct page *kvm_vcpu_nopage(struct vm_area_struct *vma, |
752 | unsigned long address, | |
753 | int *type) | |
754 | { | |
755 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
756 | unsigned long pgoff; | |
757 | struct page *page; | |
758 | ||
9a2bb7f4 | 759 | pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
039576c0 AK |
760 | if (pgoff == 0) |
761 | page = virt_to_page(vcpu->run); | |
762 | else if (pgoff == KVM_PIO_PAGE_OFFSET) | |
763 | page = virt_to_page(vcpu->pio_data); | |
764 | else | |
9a2bb7f4 | 765 | return NOPAGE_SIGBUS; |
9a2bb7f4 | 766 | get_page(page); |
cd0d9137 NAQ |
767 | if (type != NULL) |
768 | *type = VM_FAULT_MINOR; | |
769 | ||
9a2bb7f4 AK |
770 | return page; |
771 | } | |
772 | ||
773 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
774 | .nopage = kvm_vcpu_nopage, | |
775 | }; | |
776 | ||
777 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
778 | { | |
779 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
780 | return 0; | |
781 | } | |
782 | ||
bccf2150 AK |
783 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
784 | { | |
785 | struct kvm_vcpu *vcpu = filp->private_data; | |
786 | ||
787 | fput(vcpu->kvm->filp); | |
788 | return 0; | |
789 | } | |
790 | ||
791 | static struct file_operations kvm_vcpu_fops = { | |
792 | .release = kvm_vcpu_release, | |
793 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
794 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 795 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
796 | }; |
797 | ||
798 | /* | |
799 | * Allocates an inode for the vcpu. | |
800 | */ | |
801 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
802 | { | |
803 | int fd, r; | |
804 | struct inode *inode; | |
805 | struct file *file; | |
806 | ||
d6d28168 AK |
807 | r = anon_inode_getfd(&fd, &inode, &file, |
808 | "kvm-vcpu", &kvm_vcpu_fops, vcpu); | |
809 | if (r) | |
810 | return r; | |
bccf2150 | 811 | atomic_inc(&vcpu->kvm->filp->f_count); |
bccf2150 | 812 | return fd; |
bccf2150 AK |
813 | } |
814 | ||
c5ea7660 AK |
815 | /* |
816 | * Creates some virtual cpus. Good luck creating more than one. | |
817 | */ | |
818 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
819 | { | |
820 | int r; | |
821 | struct kvm_vcpu *vcpu; | |
822 | ||
c5ea7660 | 823 | if (!valid_vcpu(n)) |
fb3f0f51 | 824 | return -EINVAL; |
c5ea7660 | 825 | |
e9b11c17 | 826 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
827 | if (IS_ERR(vcpu)) |
828 | return PTR_ERR(vcpu); | |
c5ea7660 | 829 | |
15ad7146 AK |
830 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
831 | ||
11ec2804 | 832 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
833 | if (kvm->vcpus[n]) { |
834 | r = -EEXIST; | |
11ec2804 | 835 | mutex_unlock(&kvm->lock); |
e9b11c17 | 836 | goto vcpu_destroy; |
fb3f0f51 RR |
837 | } |
838 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 839 | mutex_unlock(&kvm->lock); |
c5ea7660 | 840 | |
fb3f0f51 | 841 | /* Now it's all set up, let userspace reach it */ |
bccf2150 AK |
842 | r = create_vcpu_fd(vcpu); |
843 | if (r < 0) | |
fb3f0f51 RR |
844 | goto unlink; |
845 | return r; | |
39c3b86e | 846 | |
fb3f0f51 | 847 | unlink: |
11ec2804 | 848 | mutex_lock(&kvm->lock); |
fb3f0f51 | 849 | kvm->vcpus[n] = NULL; |
11ec2804 | 850 | mutex_unlock(&kvm->lock); |
e9b11c17 ZX |
851 | vcpu_destroy: |
852 | kvm_arch_vcpu_destory(vcpu); | |
c5ea7660 AK |
853 | return r; |
854 | } | |
855 | ||
1961d276 AK |
856 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
857 | { | |
858 | if (sigset) { | |
859 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
860 | vcpu->sigset_active = 1; | |
861 | vcpu->sigset = *sigset; | |
862 | } else | |
863 | vcpu->sigset_active = 0; | |
864 | return 0; | |
865 | } | |
866 | ||
bccf2150 AK |
867 | static long kvm_vcpu_ioctl(struct file *filp, |
868 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 869 | { |
bccf2150 | 870 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 871 | void __user *argp = (void __user *)arg; |
313a3dc7 | 872 | int r; |
6aa8b732 AK |
873 | |
874 | switch (ioctl) { | |
9a2bb7f4 | 875 | case KVM_RUN: |
f0fe5108 AK |
876 | r = -EINVAL; |
877 | if (arg) | |
878 | goto out; | |
b6c7a5dc | 879 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 880 | break; |
6aa8b732 AK |
881 | case KVM_GET_REGS: { |
882 | struct kvm_regs kvm_regs; | |
883 | ||
bccf2150 | 884 | memset(&kvm_regs, 0, sizeof kvm_regs); |
b6c7a5dc | 885 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, &kvm_regs); |
6aa8b732 AK |
886 | if (r) |
887 | goto out; | |
888 | r = -EFAULT; | |
2f366987 | 889 | if (copy_to_user(argp, &kvm_regs, sizeof kvm_regs)) |
6aa8b732 AK |
890 | goto out; |
891 | r = 0; | |
892 | break; | |
893 | } | |
894 | case KVM_SET_REGS: { | |
895 | struct kvm_regs kvm_regs; | |
896 | ||
897 | r = -EFAULT; | |
2f366987 | 898 | if (copy_from_user(&kvm_regs, argp, sizeof kvm_regs)) |
6aa8b732 | 899 | goto out; |
b6c7a5dc | 900 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, &kvm_regs); |
6aa8b732 AK |
901 | if (r) |
902 | goto out; | |
903 | r = 0; | |
904 | break; | |
905 | } | |
906 | case KVM_GET_SREGS: { | |
907 | struct kvm_sregs kvm_sregs; | |
908 | ||
bccf2150 | 909 | memset(&kvm_sregs, 0, sizeof kvm_sregs); |
b6c7a5dc | 910 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
911 | if (r) |
912 | goto out; | |
913 | r = -EFAULT; | |
2f366987 | 914 | if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs)) |
6aa8b732 AK |
915 | goto out; |
916 | r = 0; | |
917 | break; | |
918 | } | |
919 | case KVM_SET_SREGS: { | |
920 | struct kvm_sregs kvm_sregs; | |
921 | ||
922 | r = -EFAULT; | |
2f366987 | 923 | if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs)) |
6aa8b732 | 924 | goto out; |
b6c7a5dc | 925 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
926 | if (r) |
927 | goto out; | |
928 | r = 0; | |
929 | break; | |
930 | } | |
931 | case KVM_TRANSLATE: { | |
932 | struct kvm_translation tr; | |
933 | ||
934 | r = -EFAULT; | |
2f366987 | 935 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 936 | goto out; |
8b006791 | 937 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
938 | if (r) |
939 | goto out; | |
940 | r = -EFAULT; | |
2f366987 | 941 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
942 | goto out; |
943 | r = 0; | |
944 | break; | |
945 | } | |
946 | case KVM_INTERRUPT: { | |
947 | struct kvm_interrupt irq; | |
948 | ||
949 | r = -EFAULT; | |
2f366987 | 950 | if (copy_from_user(&irq, argp, sizeof irq)) |
6aa8b732 | 951 | goto out; |
bccf2150 | 952 | r = kvm_vcpu_ioctl_interrupt(vcpu, &irq); |
6aa8b732 AK |
953 | if (r) |
954 | goto out; | |
955 | r = 0; | |
956 | break; | |
957 | } | |
958 | case KVM_DEBUG_GUEST: { | |
959 | struct kvm_debug_guest dbg; | |
960 | ||
961 | r = -EFAULT; | |
2f366987 | 962 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 963 | goto out; |
b6c7a5dc | 964 | r = kvm_arch_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
965 | if (r) |
966 | goto out; | |
967 | r = 0; | |
968 | break; | |
969 | } | |
1961d276 AK |
970 | case KVM_SET_SIGNAL_MASK: { |
971 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
972 | struct kvm_signal_mask kvm_sigmask; | |
973 | sigset_t sigset, *p; | |
974 | ||
975 | p = NULL; | |
976 | if (argp) { | |
977 | r = -EFAULT; | |
978 | if (copy_from_user(&kvm_sigmask, argp, | |
979 | sizeof kvm_sigmask)) | |
980 | goto out; | |
981 | r = -EINVAL; | |
982 | if (kvm_sigmask.len != sizeof sigset) | |
983 | goto out; | |
984 | r = -EFAULT; | |
985 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
986 | sizeof sigset)) | |
987 | goto out; | |
988 | p = &sigset; | |
989 | } | |
990 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
991 | break; | |
992 | } | |
b8836737 AK |
993 | case KVM_GET_FPU: { |
994 | struct kvm_fpu fpu; | |
995 | ||
996 | memset(&fpu, 0, sizeof fpu); | |
d0752060 | 997 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, &fpu); |
b8836737 AK |
998 | if (r) |
999 | goto out; | |
1000 | r = -EFAULT; | |
1001 | if (copy_to_user(argp, &fpu, sizeof fpu)) | |
1002 | goto out; | |
1003 | r = 0; | |
1004 | break; | |
1005 | } | |
1006 | case KVM_SET_FPU: { | |
1007 | struct kvm_fpu fpu; | |
1008 | ||
1009 | r = -EFAULT; | |
1010 | if (copy_from_user(&fpu, argp, sizeof fpu)) | |
1011 | goto out; | |
d0752060 | 1012 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, &fpu); |
b8836737 AK |
1013 | if (r) |
1014 | goto out; | |
1015 | r = 0; | |
1016 | break; | |
1017 | } | |
bccf2150 | 1018 | default: |
313a3dc7 | 1019 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1020 | } |
1021 | out: | |
1022 | return r; | |
1023 | } | |
1024 | ||
1025 | static long kvm_vm_ioctl(struct file *filp, | |
1026 | unsigned int ioctl, unsigned long arg) | |
1027 | { | |
1028 | struct kvm *kvm = filp->private_data; | |
1029 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1030 | int r; |
bccf2150 AK |
1031 | |
1032 | switch (ioctl) { | |
1033 | case KVM_CREATE_VCPU: | |
1034 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1035 | if (r < 0) | |
1036 | goto out; | |
1037 | break; | |
6fc138d2 IE |
1038 | case KVM_SET_USER_MEMORY_REGION: { |
1039 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1040 | ||
1041 | r = -EFAULT; | |
1042 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1043 | sizeof kvm_userspace_mem)) | |
1044 | goto out; | |
1045 | ||
1046 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1047 | if (r) |
1048 | goto out; | |
1049 | break; | |
1050 | } | |
1051 | case KVM_GET_DIRTY_LOG: { | |
1052 | struct kvm_dirty_log log; | |
1053 | ||
1054 | r = -EFAULT; | |
2f366987 | 1055 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1056 | goto out; |
2c6f5df9 | 1057 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1058 | if (r) |
1059 | goto out; | |
1060 | break; | |
1061 | } | |
f17abe9a | 1062 | default: |
1fe779f8 | 1063 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
1064 | } |
1065 | out: | |
1066 | return r; | |
1067 | } | |
1068 | ||
1069 | static struct page *kvm_vm_nopage(struct vm_area_struct *vma, | |
1070 | unsigned long address, | |
1071 | int *type) | |
1072 | { | |
1073 | struct kvm *kvm = vma->vm_file->private_data; | |
1074 | unsigned long pgoff; | |
f17abe9a AK |
1075 | struct page *page; |
1076 | ||
f17abe9a | 1077 | pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
e0d62c7f IE |
1078 | if (!kvm_is_visible_gfn(kvm, pgoff)) |
1079 | return NOPAGE_SIGBUS; | |
aab61cc0 AL |
1080 | /* current->mm->mmap_sem is already held so call lockless version */ |
1081 | page = __gfn_to_page(kvm, pgoff); | |
8a7ae055 IE |
1082 | if (is_error_page(page)) { |
1083 | kvm_release_page(page); | |
f17abe9a | 1084 | return NOPAGE_SIGBUS; |
8a7ae055 | 1085 | } |
cd0d9137 NAQ |
1086 | if (type != NULL) |
1087 | *type = VM_FAULT_MINOR; | |
1088 | ||
f17abe9a AK |
1089 | return page; |
1090 | } | |
1091 | ||
1092 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
1093 | .nopage = kvm_vm_nopage, | |
1094 | }; | |
1095 | ||
1096 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1097 | { | |
1098 | vma->vm_ops = &kvm_vm_vm_ops; | |
1099 | return 0; | |
1100 | } | |
1101 | ||
1102 | static struct file_operations kvm_vm_fops = { | |
1103 | .release = kvm_vm_release, | |
1104 | .unlocked_ioctl = kvm_vm_ioctl, | |
1105 | .compat_ioctl = kvm_vm_ioctl, | |
1106 | .mmap = kvm_vm_mmap, | |
1107 | }; | |
1108 | ||
1109 | static int kvm_dev_ioctl_create_vm(void) | |
1110 | { | |
1111 | int fd, r; | |
1112 | struct inode *inode; | |
1113 | struct file *file; | |
1114 | struct kvm *kvm; | |
1115 | ||
f17abe9a | 1116 | kvm = kvm_create_vm(); |
d6d28168 AK |
1117 | if (IS_ERR(kvm)) |
1118 | return PTR_ERR(kvm); | |
1119 | r = anon_inode_getfd(&fd, &inode, &file, "kvm-vm", &kvm_vm_fops, kvm); | |
1120 | if (r) { | |
1121 | kvm_destroy_vm(kvm); | |
1122 | return r; | |
f17abe9a AK |
1123 | } |
1124 | ||
bccf2150 | 1125 | kvm->filp = file; |
f17abe9a | 1126 | |
f17abe9a | 1127 | return fd; |
f17abe9a AK |
1128 | } |
1129 | ||
1130 | static long kvm_dev_ioctl(struct file *filp, | |
1131 | unsigned int ioctl, unsigned long arg) | |
1132 | { | |
1133 | void __user *argp = (void __user *)arg; | |
07c45a36 | 1134 | long r = -EINVAL; |
f17abe9a AK |
1135 | |
1136 | switch (ioctl) { | |
1137 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1138 | r = -EINVAL; |
1139 | if (arg) | |
1140 | goto out; | |
f17abe9a AK |
1141 | r = KVM_API_VERSION; |
1142 | break; | |
1143 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1144 | r = -EINVAL; |
1145 | if (arg) | |
1146 | goto out; | |
f17abe9a AK |
1147 | r = kvm_dev_ioctl_create_vm(); |
1148 | break; | |
018d00d2 ZX |
1149 | case KVM_CHECK_EXTENSION: |
1150 | r = kvm_dev_ioctl_check_extension((long)argp); | |
5d308f45 | 1151 | break; |
07c45a36 AK |
1152 | case KVM_GET_VCPU_MMAP_SIZE: |
1153 | r = -EINVAL; | |
1154 | if (arg) | |
1155 | goto out; | |
039576c0 | 1156 | r = 2 * PAGE_SIZE; |
07c45a36 | 1157 | break; |
6aa8b732 | 1158 | default: |
043405e1 | 1159 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1160 | } |
1161 | out: | |
1162 | return r; | |
1163 | } | |
1164 | ||
6aa8b732 | 1165 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1166 | .unlocked_ioctl = kvm_dev_ioctl, |
1167 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1168 | }; |
1169 | ||
1170 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1171 | KVM_MINOR, |
6aa8b732 AK |
1172 | "kvm", |
1173 | &kvm_chardev_ops, | |
1174 | }; | |
1175 | ||
1b6c0168 AK |
1176 | static void hardware_enable(void *junk) |
1177 | { | |
1178 | int cpu = raw_smp_processor_id(); | |
1179 | ||
1180 | if (cpu_isset(cpu, cpus_hardware_enabled)) | |
1181 | return; | |
1182 | cpu_set(cpu, cpus_hardware_enabled); | |
e9b11c17 | 1183 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
1184 | } |
1185 | ||
1186 | static void hardware_disable(void *junk) | |
1187 | { | |
1188 | int cpu = raw_smp_processor_id(); | |
1189 | ||
1190 | if (!cpu_isset(cpu, cpus_hardware_enabled)) | |
1191 | return; | |
1192 | cpu_clear(cpu, cpus_hardware_enabled); | |
1193 | decache_vcpus_on_cpu(cpu); | |
e9b11c17 | 1194 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1195 | } |
1196 | ||
774c47f1 AK |
1197 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1198 | void *v) | |
1199 | { | |
1200 | int cpu = (long)v; | |
1201 | ||
1a6f4d7f | 1202 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1203 | switch (val) { |
cec9ad27 | 1204 | case CPU_DYING: |
6ec8a856 AK |
1205 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1206 | cpu); | |
1207 | hardware_disable(NULL); | |
1208 | break; | |
774c47f1 | 1209 | case CPU_UP_CANCELED: |
43934a38 JK |
1210 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1211 | cpu); | |
1b6c0168 | 1212 | smp_call_function_single(cpu, hardware_disable, NULL, 0, 1); |
774c47f1 | 1213 | break; |
43934a38 JK |
1214 | case CPU_ONLINE: |
1215 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1216 | cpu); | |
1b6c0168 | 1217 | smp_call_function_single(cpu, hardware_enable, NULL, 0, 1); |
774c47f1 AK |
1218 | break; |
1219 | } | |
1220 | return NOTIFY_OK; | |
1221 | } | |
1222 | ||
9a2b85c6 | 1223 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1224 | void *v) |
9a2b85c6 RR |
1225 | { |
1226 | if (val == SYS_RESTART) { | |
1227 | /* | |
1228 | * Some (well, at least mine) BIOSes hang on reboot if | |
1229 | * in vmx root mode. | |
1230 | */ | |
1231 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1232 | on_each_cpu(hardware_disable, NULL, 0, 1); | |
1233 | } | |
1234 | return NOTIFY_OK; | |
1235 | } | |
1236 | ||
1237 | static struct notifier_block kvm_reboot_notifier = { | |
1238 | .notifier_call = kvm_reboot, | |
1239 | .priority = 0, | |
1240 | }; | |
1241 | ||
2eeb2e94 GH |
1242 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
1243 | { | |
1244 | memset(bus, 0, sizeof(*bus)); | |
1245 | } | |
1246 | ||
1247 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
1248 | { | |
1249 | int i; | |
1250 | ||
1251 | for (i = 0; i < bus->dev_count; i++) { | |
1252 | struct kvm_io_device *pos = bus->devs[i]; | |
1253 | ||
1254 | kvm_iodevice_destructor(pos); | |
1255 | } | |
1256 | } | |
1257 | ||
1258 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr) | |
1259 | { | |
1260 | int i; | |
1261 | ||
1262 | for (i = 0; i < bus->dev_count; i++) { | |
1263 | struct kvm_io_device *pos = bus->devs[i]; | |
1264 | ||
1265 | if (pos->in_range(pos, addr)) | |
1266 | return pos; | |
1267 | } | |
1268 | ||
1269 | return NULL; | |
1270 | } | |
1271 | ||
1272 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
1273 | { | |
1274 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
1275 | ||
1276 | bus->devs[bus->dev_count++] = dev; | |
1277 | } | |
1278 | ||
774c47f1 AK |
1279 | static struct notifier_block kvm_cpu_notifier = { |
1280 | .notifier_call = kvm_cpu_hotplug, | |
1281 | .priority = 20, /* must be > scheduler priority */ | |
1282 | }; | |
1283 | ||
1165f5fe AK |
1284 | static u64 stat_get(void *_offset) |
1285 | { | |
1286 | unsigned offset = (long)_offset; | |
1287 | u64 total = 0; | |
1288 | struct kvm *kvm; | |
1289 | struct kvm_vcpu *vcpu; | |
1290 | int i; | |
1291 | ||
1292 | spin_lock(&kvm_lock); | |
1293 | list_for_each_entry(kvm, &vm_list, vm_list) | |
1294 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
1295 | vcpu = kvm->vcpus[i]; |
1296 | if (vcpu) | |
1297 | total += *(u32 *)((void *)vcpu + offset); | |
1165f5fe AK |
1298 | } |
1299 | spin_unlock(&kvm_lock); | |
1300 | return total; | |
1301 | } | |
1302 | ||
3dea7ca7 | 1303 | DEFINE_SIMPLE_ATTRIBUTE(stat_fops, stat_get, NULL, "%llu\n"); |
1165f5fe | 1304 | |
a16b043c | 1305 | static void kvm_init_debug(void) |
6aa8b732 AK |
1306 | { |
1307 | struct kvm_stats_debugfs_item *p; | |
1308 | ||
8b6d44c7 | 1309 | debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 1310 | for (p = debugfs_entries; p->name; ++p) |
1165f5fe AK |
1311 | p->dentry = debugfs_create_file(p->name, 0444, debugfs_dir, |
1312 | (void *)(long)p->offset, | |
1313 | &stat_fops); | |
6aa8b732 AK |
1314 | } |
1315 | ||
1316 | static void kvm_exit_debug(void) | |
1317 | { | |
1318 | struct kvm_stats_debugfs_item *p; | |
1319 | ||
1320 | for (p = debugfs_entries; p->name; ++p) | |
1321 | debugfs_remove(p->dentry); | |
1322 | debugfs_remove(debugfs_dir); | |
1323 | } | |
1324 | ||
59ae6c6b AK |
1325 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
1326 | { | |
4267c41a | 1327 | hardware_disable(NULL); |
59ae6c6b AK |
1328 | return 0; |
1329 | } | |
1330 | ||
1331 | static int kvm_resume(struct sys_device *dev) | |
1332 | { | |
4267c41a | 1333 | hardware_enable(NULL); |
59ae6c6b AK |
1334 | return 0; |
1335 | } | |
1336 | ||
1337 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 1338 | .name = "kvm", |
59ae6c6b AK |
1339 | .suspend = kvm_suspend, |
1340 | .resume = kvm_resume, | |
1341 | }; | |
1342 | ||
1343 | static struct sys_device kvm_sysdev = { | |
1344 | .id = 0, | |
1345 | .cls = &kvm_sysdev_class, | |
1346 | }; | |
1347 | ||
cea7bb21 | 1348 | struct page *bad_page; |
6aa8b732 | 1349 | |
15ad7146 AK |
1350 | static inline |
1351 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
1352 | { | |
1353 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
1354 | } | |
1355 | ||
1356 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
1357 | { | |
1358 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1359 | ||
e9b11c17 | 1360 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
1361 | } |
1362 | ||
1363 | static void kvm_sched_out(struct preempt_notifier *pn, | |
1364 | struct task_struct *next) | |
1365 | { | |
1366 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1367 | ||
e9b11c17 | 1368 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
1369 | } |
1370 | ||
f8c16bba | 1371 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 1372 | struct module *module) |
6aa8b732 AK |
1373 | { |
1374 | int r; | |
002c7f7c | 1375 | int cpu; |
6aa8b732 | 1376 | |
cb498ea2 ZX |
1377 | r = kvm_mmu_module_init(); |
1378 | if (r) | |
1379 | goto out4; | |
1380 | ||
1381 | kvm_init_debug(); | |
1382 | ||
f8c16bba ZX |
1383 | r = kvm_arch_init(opaque); |
1384 | if (r) | |
1385 | goto out4; | |
cb498ea2 ZX |
1386 | |
1387 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
1388 | ||
1389 | if (bad_page == NULL) { | |
1390 | r = -ENOMEM; | |
1391 | goto out; | |
1392 | } | |
1393 | ||
e9b11c17 | 1394 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 1395 | if (r < 0) |
ca45aaae | 1396 | goto out; |
6aa8b732 | 1397 | |
002c7f7c YS |
1398 | for_each_online_cpu(cpu) { |
1399 | smp_call_function_single(cpu, | |
e9b11c17 | 1400 | kvm_arch_check_processor_compat, |
002c7f7c YS |
1401 | &r, 0, 1); |
1402 | if (r < 0) | |
1403 | goto out_free_0; | |
1404 | } | |
1405 | ||
1b6c0168 | 1406 | on_each_cpu(hardware_enable, NULL, 0, 1); |
774c47f1 AK |
1407 | r = register_cpu_notifier(&kvm_cpu_notifier); |
1408 | if (r) | |
1409 | goto out_free_1; | |
6aa8b732 AK |
1410 | register_reboot_notifier(&kvm_reboot_notifier); |
1411 | ||
59ae6c6b AK |
1412 | r = sysdev_class_register(&kvm_sysdev_class); |
1413 | if (r) | |
1414 | goto out_free_2; | |
1415 | ||
1416 | r = sysdev_register(&kvm_sysdev); | |
1417 | if (r) | |
1418 | goto out_free_3; | |
1419 | ||
c16f862d RR |
1420 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
1421 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
1422 | __alignof__(struct kvm_vcpu), |
1423 | 0, NULL); | |
c16f862d RR |
1424 | if (!kvm_vcpu_cache) { |
1425 | r = -ENOMEM; | |
1426 | goto out_free_4; | |
1427 | } | |
1428 | ||
6aa8b732 AK |
1429 | kvm_chardev_ops.owner = module; |
1430 | ||
1431 | r = misc_register(&kvm_dev); | |
1432 | if (r) { | |
d77c26fc | 1433 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
1434 | goto out_free; |
1435 | } | |
1436 | ||
15ad7146 AK |
1437 | kvm_preempt_ops.sched_in = kvm_sched_in; |
1438 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
1439 | ||
c7addb90 AK |
1440 | kvm_mmu_set_nonpresent_ptes(0ull, 0ull); |
1441 | ||
1442 | return 0; | |
6aa8b732 AK |
1443 | |
1444 | out_free: | |
c16f862d RR |
1445 | kmem_cache_destroy(kvm_vcpu_cache); |
1446 | out_free_4: | |
59ae6c6b AK |
1447 | sysdev_unregister(&kvm_sysdev); |
1448 | out_free_3: | |
1449 | sysdev_class_unregister(&kvm_sysdev_class); | |
1450 | out_free_2: | |
6aa8b732 | 1451 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 AK |
1452 | unregister_cpu_notifier(&kvm_cpu_notifier); |
1453 | out_free_1: | |
1b6c0168 | 1454 | on_each_cpu(hardware_disable, NULL, 0, 1); |
002c7f7c | 1455 | out_free_0: |
e9b11c17 | 1456 | kvm_arch_hardware_unsetup(); |
ca45aaae | 1457 | out: |
f8c16bba | 1458 | kvm_arch_exit(); |
cb498ea2 ZX |
1459 | kvm_exit_debug(); |
1460 | kvm_mmu_module_exit(); | |
1461 | out4: | |
6aa8b732 AK |
1462 | return r; |
1463 | } | |
cb498ea2 | 1464 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 1465 | |
cb498ea2 | 1466 | void kvm_exit(void) |
6aa8b732 AK |
1467 | { |
1468 | misc_deregister(&kvm_dev); | |
c16f862d | 1469 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
1470 | sysdev_unregister(&kvm_sysdev); |
1471 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 1472 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 1473 | unregister_cpu_notifier(&kvm_cpu_notifier); |
1b6c0168 | 1474 | on_each_cpu(hardware_disable, NULL, 0, 1); |
e9b11c17 | 1475 | kvm_arch_hardware_unsetup(); |
f8c16bba | 1476 | kvm_arch_exit(); |
6aa8b732 | 1477 | kvm_exit_debug(); |
cea7bb21 | 1478 | __free_page(bad_page); |
b5a33a75 | 1479 | kvm_mmu_module_exit(); |
6aa8b732 | 1480 | } |
cb498ea2 | 1481 | EXPORT_SYMBOL_GPL(kvm_exit); |