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> |
6aa8b732 | 43 | |
e495606d | 44 | #include <asm/processor.h> |
e495606d AK |
45 | #include <asm/io.h> |
46 | #include <asm/uaccess.h> | |
3e021bf5 | 47 | #include <asm/pgtable.h> |
6aa8b732 AK |
48 | |
49 | MODULE_AUTHOR("Qumranet"); | |
50 | MODULE_LICENSE("GPL"); | |
51 | ||
e9b11c17 ZX |
52 | DEFINE_SPINLOCK(kvm_lock); |
53 | LIST_HEAD(vm_list); | |
133de902 | 54 | |
1b6c0168 AK |
55 | static cpumask_t cpus_hardware_enabled; |
56 | ||
c16f862d RR |
57 | struct kmem_cache *kvm_vcpu_cache; |
58 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 59 | |
15ad7146 AK |
60 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
61 | ||
6aa8b732 AK |
62 | static struct dentry *debugfs_dir; |
63 | ||
bccf2150 AK |
64 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
65 | unsigned long arg); | |
66 | ||
5aacf0ca JM |
67 | static inline int valid_vcpu(int n) |
68 | { | |
69 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
70 | } | |
71 | ||
bccf2150 AK |
72 | /* |
73 | * Switches to specified vcpu, until a matching vcpu_put() | |
74 | */ | |
313a3dc7 | 75 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 76 | { |
15ad7146 AK |
77 | int cpu; |
78 | ||
bccf2150 | 79 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
80 | cpu = get_cpu(); |
81 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 82 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 83 | put_cpu(); |
6aa8b732 AK |
84 | } |
85 | ||
313a3dc7 | 86 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 87 | { |
15ad7146 | 88 | preempt_disable(); |
313a3dc7 | 89 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
90 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
91 | preempt_enable(); | |
6aa8b732 AK |
92 | mutex_unlock(&vcpu->mutex); |
93 | } | |
94 | ||
d9e368d6 AK |
95 | static void ack_flush(void *_completed) |
96 | { | |
d9e368d6 AK |
97 | } |
98 | ||
99 | void kvm_flush_remote_tlbs(struct kvm *kvm) | |
100 | { | |
49d3bd7e | 101 | int i, cpu; |
d9e368d6 AK |
102 | cpumask_t cpus; |
103 | struct kvm_vcpu *vcpu; | |
d9e368d6 | 104 | |
d9e368d6 | 105 | cpus_clear(cpus); |
fb3f0f51 RR |
106 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
107 | vcpu = kvm->vcpus[i]; | |
108 | if (!vcpu) | |
109 | continue; | |
3176bc3e | 110 | if (test_and_set_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests)) |
d9e368d6 AK |
111 | continue; |
112 | cpu = vcpu->cpu; | |
113 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
49d3bd7e | 114 | cpu_set(cpu, cpus); |
d9e368d6 | 115 | } |
0f74a24c AK |
116 | if (cpus_empty(cpus)) |
117 | return; | |
118 | ++kvm->stat.remote_tlb_flush; | |
49d3bd7e | 119 | smp_call_function_mask(cpus, ack_flush, NULL, 1); |
d9e368d6 AK |
120 | } |
121 | ||
fb3f0f51 RR |
122 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
123 | { | |
124 | struct page *page; | |
125 | int r; | |
126 | ||
127 | mutex_init(&vcpu->mutex); | |
128 | vcpu->cpu = -1; | |
fb3f0f51 RR |
129 | vcpu->kvm = kvm; |
130 | vcpu->vcpu_id = id; | |
b6958ce4 | 131 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
132 | |
133 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
134 | if (!page) { | |
135 | r = -ENOMEM; | |
136 | goto fail; | |
137 | } | |
138 | vcpu->run = page_address(page); | |
139 | ||
e9b11c17 | 140 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 141 | if (r < 0) |
e9b11c17 | 142 | goto fail_free_run; |
fb3f0f51 RR |
143 | return 0; |
144 | ||
fb3f0f51 RR |
145 | fail_free_run: |
146 | free_page((unsigned long)vcpu->run); | |
147 | fail: | |
76fafa5e | 148 | return r; |
fb3f0f51 RR |
149 | } |
150 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
151 | ||
152 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
153 | { | |
e9b11c17 | 154 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
155 | free_page((unsigned long)vcpu->run); |
156 | } | |
157 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
158 | ||
f17abe9a | 159 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 160 | { |
d19a9cd2 | 161 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 162 | |
d19a9cd2 ZX |
163 | if (IS_ERR(kvm)) |
164 | goto out; | |
6aa8b732 | 165 | |
6d4e4c4f AK |
166 | kvm->mm = current->mm; |
167 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 168 | spin_lock_init(&kvm->mmu_lock); |
74906345 | 169 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 170 | mutex_init(&kvm->lock); |
2eeb2e94 | 171 | kvm_io_bus_init(&kvm->mmio_bus); |
72dc67a6 | 172 | init_rwsem(&kvm->slots_lock); |
5e58cfe4 RR |
173 | spin_lock(&kvm_lock); |
174 | list_add(&kvm->vm_list, &vm_list); | |
175 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 176 | out: |
f17abe9a AK |
177 | return kvm; |
178 | } | |
179 | ||
6aa8b732 AK |
180 | /* |
181 | * Free any memory in @free but not in @dont. | |
182 | */ | |
183 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
184 | struct kvm_memory_slot *dont) | |
185 | { | |
290fc38d IE |
186 | if (!dont || free->rmap != dont->rmap) |
187 | vfree(free->rmap); | |
6aa8b732 AK |
188 | |
189 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
190 | vfree(free->dirty_bitmap); | |
191 | ||
6aa8b732 | 192 | free->npages = 0; |
8b6d44c7 | 193 | free->dirty_bitmap = NULL; |
8d4e1288 | 194 | free->rmap = NULL; |
6aa8b732 AK |
195 | } |
196 | ||
d19a9cd2 | 197 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
198 | { |
199 | int i; | |
200 | ||
201 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 202 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
203 | } |
204 | ||
f17abe9a AK |
205 | static void kvm_destroy_vm(struct kvm *kvm) |
206 | { | |
6d4e4c4f AK |
207 | struct mm_struct *mm = kvm->mm; |
208 | ||
133de902 AK |
209 | spin_lock(&kvm_lock); |
210 | list_del(&kvm->vm_list); | |
211 | spin_unlock(&kvm_lock); | |
74906345 | 212 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 213 | kvm_io_bus_destroy(&kvm->mmio_bus); |
d19a9cd2 | 214 | kvm_arch_destroy_vm(kvm); |
6d4e4c4f | 215 | mmdrop(mm); |
f17abe9a AK |
216 | } |
217 | ||
218 | static int kvm_vm_release(struct inode *inode, struct file *filp) | |
219 | { | |
220 | struct kvm *kvm = filp->private_data; | |
221 | ||
222 | kvm_destroy_vm(kvm); | |
6aa8b732 AK |
223 | return 0; |
224 | } | |
225 | ||
6aa8b732 AK |
226 | /* |
227 | * Allocate some memory and give it an address in the guest physical address | |
228 | * space. | |
229 | * | |
230 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 231 | * |
10589a46 | 232 | * Must be called holding mmap_sem for write. |
6aa8b732 | 233 | */ |
f78e0e2e SY |
234 | int __kvm_set_memory_region(struct kvm *kvm, |
235 | struct kvm_userspace_memory_region *mem, | |
236 | int user_alloc) | |
6aa8b732 AK |
237 | { |
238 | int r; | |
239 | gfn_t base_gfn; | |
240 | unsigned long npages; | |
241 | unsigned long i; | |
242 | struct kvm_memory_slot *memslot; | |
243 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
244 | |
245 | r = -EINVAL; | |
246 | /* General sanity checks */ | |
247 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
248 | goto out; | |
249 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
250 | goto out; | |
e0d62c7f | 251 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
252 | goto out; |
253 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
254 | goto out; | |
255 | ||
256 | memslot = &kvm->memslots[mem->slot]; | |
257 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
258 | npages = mem->memory_size >> PAGE_SHIFT; | |
259 | ||
260 | if (!npages) | |
261 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
262 | ||
6aa8b732 AK |
263 | new = old = *memslot; |
264 | ||
265 | new.base_gfn = base_gfn; | |
266 | new.npages = npages; | |
267 | new.flags = mem->flags; | |
268 | ||
269 | /* Disallow changing a memory slot's size. */ | |
270 | r = -EINVAL; | |
271 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 272 | goto out_free; |
6aa8b732 AK |
273 | |
274 | /* Check for overlaps */ | |
275 | r = -EEXIST; | |
276 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
277 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
278 | ||
279 | if (s == memslot) | |
280 | continue; | |
281 | if (!((base_gfn + npages <= s->base_gfn) || | |
282 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 283 | goto out_free; |
6aa8b732 | 284 | } |
6aa8b732 | 285 | |
6aa8b732 AK |
286 | /* Free page dirty bitmap if unneeded */ |
287 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 288 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
289 | |
290 | r = -ENOMEM; | |
291 | ||
292 | /* Allocate if a slot is being created */ | |
8d4e1288 | 293 | if (npages && !new.rmap) { |
d77c26fc | 294 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
295 | |
296 | if (!new.rmap) | |
f78e0e2e | 297 | goto out_free; |
290fc38d | 298 | |
290fc38d | 299 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 300 | |
80b14b5b | 301 | new.user_alloc = user_alloc; |
0de10343 | 302 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 AK |
303 | } |
304 | ||
305 | /* Allocate page dirty bitmap if needed */ | |
306 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
307 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
308 | ||
309 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
310 | if (!new.dirty_bitmap) | |
f78e0e2e | 311 | goto out_free; |
6aa8b732 AK |
312 | memset(new.dirty_bitmap, 0, dirty_bytes); |
313 | } | |
314 | ||
6aa8b732 AK |
315 | if (mem->slot >= kvm->nmemslots) |
316 | kvm->nmemslots = mem->slot + 1; | |
317 | ||
3ad82a7e ZX |
318 | *memslot = new; |
319 | ||
0de10343 ZX |
320 | r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc); |
321 | if (r) { | |
322 | *memslot = old; | |
323 | goto out_free; | |
82ce2c96 IE |
324 | } |
325 | ||
6aa8b732 AK |
326 | kvm_free_physmem_slot(&old, &new); |
327 | return 0; | |
328 | ||
f78e0e2e | 329 | out_free: |
6aa8b732 AK |
330 | kvm_free_physmem_slot(&new, &old); |
331 | out: | |
332 | return r; | |
210c7c4d IE |
333 | |
334 | } | |
f78e0e2e SY |
335 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
336 | ||
337 | int kvm_set_memory_region(struct kvm *kvm, | |
338 | struct kvm_userspace_memory_region *mem, | |
339 | int user_alloc) | |
340 | { | |
341 | int r; | |
342 | ||
72dc67a6 | 343 | down_write(&kvm->slots_lock); |
f78e0e2e | 344 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
72dc67a6 | 345 | up_write(&kvm->slots_lock); |
f78e0e2e SY |
346 | return r; |
347 | } | |
210c7c4d IE |
348 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
349 | ||
1fe779f8 CO |
350 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
351 | struct | |
352 | kvm_userspace_memory_region *mem, | |
353 | int user_alloc) | |
210c7c4d | 354 | { |
e0d62c7f IE |
355 | if (mem->slot >= KVM_MEMORY_SLOTS) |
356 | return -EINVAL; | |
210c7c4d | 357 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
358 | } |
359 | ||
5bb064dc ZX |
360 | int kvm_get_dirty_log(struct kvm *kvm, |
361 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
362 | { |
363 | struct kvm_memory_slot *memslot; | |
364 | int r, i; | |
365 | int n; | |
366 | unsigned long any = 0; | |
367 | ||
6aa8b732 AK |
368 | r = -EINVAL; |
369 | if (log->slot >= KVM_MEMORY_SLOTS) | |
370 | goto out; | |
371 | ||
372 | memslot = &kvm->memslots[log->slot]; | |
373 | r = -ENOENT; | |
374 | if (!memslot->dirty_bitmap) | |
375 | goto out; | |
376 | ||
cd1a4a98 | 377 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 378 | |
cd1a4a98 | 379 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
380 | any = memslot->dirty_bitmap[i]; |
381 | ||
382 | r = -EFAULT; | |
383 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
384 | goto out; | |
385 | ||
5bb064dc ZX |
386 | if (any) |
387 | *is_dirty = 1; | |
6aa8b732 AK |
388 | |
389 | r = 0; | |
6aa8b732 | 390 | out: |
6aa8b732 AK |
391 | return r; |
392 | } | |
393 | ||
cea7bb21 IE |
394 | int is_error_page(struct page *page) |
395 | { | |
396 | return page == bad_page; | |
397 | } | |
398 | EXPORT_SYMBOL_GPL(is_error_page); | |
399 | ||
f9d46eb0 IE |
400 | static inline unsigned long bad_hva(void) |
401 | { | |
402 | return PAGE_OFFSET; | |
403 | } | |
404 | ||
405 | int kvm_is_error_hva(unsigned long addr) | |
406 | { | |
407 | return addr == bad_hva(); | |
408 | } | |
409 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
410 | ||
e8207547 | 411 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
412 | { |
413 | int i; | |
414 | ||
415 | for (i = 0; i < kvm->nmemslots; ++i) { | |
416 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
417 | ||
418 | if (gfn >= memslot->base_gfn | |
419 | && gfn < memslot->base_gfn + memslot->npages) | |
420 | return memslot; | |
421 | } | |
8b6d44c7 | 422 | return NULL; |
6aa8b732 | 423 | } |
e8207547 AK |
424 | |
425 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
426 | { | |
427 | gfn = unalias_gfn(kvm, gfn); | |
428 | return __gfn_to_memslot(kvm, gfn); | |
429 | } | |
6aa8b732 | 430 | |
e0d62c7f IE |
431 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
432 | { | |
433 | int i; | |
434 | ||
435 | gfn = unalias_gfn(kvm, gfn); | |
436 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
437 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
438 | ||
439 | if (gfn >= memslot->base_gfn | |
440 | && gfn < memslot->base_gfn + memslot->npages) | |
441 | return 1; | |
442 | } | |
443 | return 0; | |
444 | } | |
445 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
446 | ||
539cb660 IE |
447 | static unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
448 | { | |
449 | struct kvm_memory_slot *slot; | |
450 | ||
451 | gfn = unalias_gfn(kvm, gfn); | |
452 | slot = __gfn_to_memslot(kvm, gfn); | |
453 | if (!slot) | |
454 | return bad_hva(); | |
455 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
456 | } | |
457 | ||
aab61cc0 AL |
458 | /* |
459 | * Requires current->mm->mmap_sem to be held | |
460 | */ | |
10589a46 | 461 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
954bbbc2 | 462 | { |
8d4e1288 | 463 | struct page *page[1]; |
539cb660 | 464 | unsigned long addr; |
8d4e1288 | 465 | int npages; |
954bbbc2 | 466 | |
60395224 AK |
467 | might_sleep(); |
468 | ||
539cb660 IE |
469 | addr = gfn_to_hva(kvm, gfn); |
470 | if (kvm_is_error_hva(addr)) { | |
8a7ae055 | 471 | get_page(bad_page); |
cea7bb21 | 472 | return bad_page; |
8a7ae055 | 473 | } |
8d4e1288 | 474 | |
539cb660 IE |
475 | npages = get_user_pages(current, current->mm, addr, 1, 1, 1, page, |
476 | NULL); | |
477 | ||
8d4e1288 AL |
478 | if (npages != 1) { |
479 | get_page(bad_page); | |
480 | return bad_page; | |
8a7ae055 | 481 | } |
8d4e1288 AL |
482 | |
483 | return page[0]; | |
954bbbc2 | 484 | } |
aab61cc0 | 485 | |
954bbbc2 AK |
486 | EXPORT_SYMBOL_GPL(gfn_to_page); |
487 | ||
b4231d61 IE |
488 | void kvm_release_page_clean(struct page *page) |
489 | { | |
490 | put_page(page); | |
491 | } | |
492 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
493 | ||
494 | void kvm_release_page_dirty(struct page *page) | |
8a7ae055 IE |
495 | { |
496 | if (!PageReserved(page)) | |
497 | SetPageDirty(page); | |
498 | put_page(page); | |
499 | } | |
b4231d61 | 500 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); |
8a7ae055 | 501 | |
195aefde IE |
502 | static int next_segment(unsigned long len, int offset) |
503 | { | |
504 | if (len > PAGE_SIZE - offset) | |
505 | return PAGE_SIZE - offset; | |
506 | else | |
507 | return len; | |
508 | } | |
509 | ||
510 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
511 | int len) | |
512 | { | |
e0506bcb IE |
513 | int r; |
514 | unsigned long addr; | |
195aefde | 515 | |
e0506bcb IE |
516 | addr = gfn_to_hva(kvm, gfn); |
517 | if (kvm_is_error_hva(addr)) | |
518 | return -EFAULT; | |
519 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
520 | if (r) | |
195aefde | 521 | return -EFAULT; |
195aefde IE |
522 | return 0; |
523 | } | |
524 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
525 | ||
526 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
527 | { | |
528 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
529 | int seg; | |
530 | int offset = offset_in_page(gpa); | |
531 | int ret; | |
532 | ||
533 | while ((seg = next_segment(len, offset)) != 0) { | |
534 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
535 | if (ret < 0) | |
536 | return ret; | |
537 | offset = 0; | |
538 | len -= seg; | |
539 | data += seg; | |
540 | ++gfn; | |
541 | } | |
542 | return 0; | |
543 | } | |
544 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
545 | ||
7ec54588 MT |
546 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
547 | unsigned long len) | |
548 | { | |
549 | int r; | |
550 | unsigned long addr; | |
551 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
552 | int offset = offset_in_page(gpa); | |
553 | ||
554 | addr = gfn_to_hva(kvm, gfn); | |
555 | if (kvm_is_error_hva(addr)) | |
556 | return -EFAULT; | |
7ec54588 | 557 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
7ec54588 MT |
558 | if (r) |
559 | return -EFAULT; | |
560 | return 0; | |
561 | } | |
562 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
563 | ||
195aefde IE |
564 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
565 | int offset, int len) | |
566 | { | |
e0506bcb IE |
567 | int r; |
568 | unsigned long addr; | |
195aefde | 569 | |
e0506bcb IE |
570 | addr = gfn_to_hva(kvm, gfn); |
571 | if (kvm_is_error_hva(addr)) | |
572 | return -EFAULT; | |
573 | r = copy_to_user((void __user *)addr + offset, data, len); | |
574 | if (r) | |
195aefde | 575 | return -EFAULT; |
195aefde IE |
576 | mark_page_dirty(kvm, gfn); |
577 | return 0; | |
578 | } | |
579 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
580 | ||
581 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
582 | unsigned long len) | |
583 | { | |
584 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
585 | int seg; | |
586 | int offset = offset_in_page(gpa); | |
587 | int ret; | |
588 | ||
589 | while ((seg = next_segment(len, offset)) != 0) { | |
590 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
591 | if (ret < 0) | |
592 | return ret; | |
593 | offset = 0; | |
594 | len -= seg; | |
595 | data += seg; | |
596 | ++gfn; | |
597 | } | |
598 | return 0; | |
599 | } | |
600 | ||
601 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
602 | { | |
3e021bf5 | 603 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
604 | } |
605 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
606 | ||
607 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
608 | { | |
609 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
610 | int seg; | |
611 | int offset = offset_in_page(gpa); | |
612 | int ret; | |
613 | ||
614 | while ((seg = next_segment(len, offset)) != 0) { | |
615 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
616 | if (ret < 0) | |
617 | return ret; | |
618 | offset = 0; | |
619 | len -= seg; | |
620 | ++gfn; | |
621 | } | |
622 | return 0; | |
623 | } | |
624 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
625 | ||
6aa8b732 AK |
626 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
627 | { | |
31389947 | 628 | struct kvm_memory_slot *memslot; |
6aa8b732 | 629 | |
3b6fff19 | 630 | gfn = unalias_gfn(kvm, gfn); |
7e9d619d RR |
631 | memslot = __gfn_to_memslot(kvm, gfn); |
632 | if (memslot && memslot->dirty_bitmap) { | |
633 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 634 | |
7e9d619d RR |
635 | /* avoid RMW */ |
636 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
637 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
638 | } |
639 | } | |
640 | ||
b6958ce4 ED |
641 | /* |
642 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
643 | */ | |
8776e519 | 644 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 645 | { |
b6958ce4 ED |
646 | DECLARE_WAITQUEUE(wait, current); |
647 | ||
648 | add_wait_queue(&vcpu->wq, &wait); | |
649 | ||
650 | /* | |
651 | * We will block until either an interrupt or a signal wakes us up | |
652 | */ | |
c5ec1534 HQ |
653 | while (!kvm_cpu_has_interrupt(vcpu) |
654 | && !signal_pending(current) | |
53e0aa7b | 655 | && !kvm_arch_vcpu_runnable(vcpu)) { |
b6958ce4 ED |
656 | set_current_state(TASK_INTERRUPTIBLE); |
657 | vcpu_put(vcpu); | |
658 | schedule(); | |
659 | vcpu_load(vcpu); | |
660 | } | |
d3bef15f | 661 | |
c5ec1534 | 662 | __set_current_state(TASK_RUNNING); |
b6958ce4 | 663 | remove_wait_queue(&vcpu->wq, &wait); |
b6958ce4 ED |
664 | } |
665 | ||
6aa8b732 AK |
666 | void kvm_resched(struct kvm_vcpu *vcpu) |
667 | { | |
3fca0365 YD |
668 | if (!need_resched()) |
669 | return; | |
6aa8b732 | 670 | cond_resched(); |
6aa8b732 AK |
671 | } |
672 | EXPORT_SYMBOL_GPL(kvm_resched); | |
673 | ||
e4a533a4 | 674 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
675 | { |
676 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
677 | struct page *page; |
678 | ||
e4a533a4 | 679 | if (vmf->pgoff == 0) |
039576c0 | 680 | page = virt_to_page(vcpu->run); |
e4a533a4 | 681 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 682 | page = virt_to_page(vcpu->arch.pio_data); |
039576c0 | 683 | else |
e4a533a4 | 684 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 685 | get_page(page); |
e4a533a4 | 686 | vmf->page = page; |
687 | return 0; | |
9a2bb7f4 AK |
688 | } |
689 | ||
690 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
e4a533a4 | 691 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
692 | }; |
693 | ||
694 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
695 | { | |
696 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
697 | return 0; | |
698 | } | |
699 | ||
bccf2150 AK |
700 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
701 | { | |
702 | struct kvm_vcpu *vcpu = filp->private_data; | |
703 | ||
704 | fput(vcpu->kvm->filp); | |
705 | return 0; | |
706 | } | |
707 | ||
708 | static struct file_operations kvm_vcpu_fops = { | |
709 | .release = kvm_vcpu_release, | |
710 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
711 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 712 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
713 | }; |
714 | ||
715 | /* | |
716 | * Allocates an inode for the vcpu. | |
717 | */ | |
718 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
719 | { | |
720 | int fd, r; | |
721 | struct inode *inode; | |
722 | struct file *file; | |
723 | ||
d6d28168 AK |
724 | r = anon_inode_getfd(&fd, &inode, &file, |
725 | "kvm-vcpu", &kvm_vcpu_fops, vcpu); | |
726 | if (r) | |
727 | return r; | |
bccf2150 | 728 | atomic_inc(&vcpu->kvm->filp->f_count); |
bccf2150 | 729 | return fd; |
bccf2150 AK |
730 | } |
731 | ||
c5ea7660 AK |
732 | /* |
733 | * Creates some virtual cpus. Good luck creating more than one. | |
734 | */ | |
735 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
736 | { | |
737 | int r; | |
738 | struct kvm_vcpu *vcpu; | |
739 | ||
c5ea7660 | 740 | if (!valid_vcpu(n)) |
fb3f0f51 | 741 | return -EINVAL; |
c5ea7660 | 742 | |
e9b11c17 | 743 | vcpu = kvm_arch_vcpu_create(kvm, n); |
fb3f0f51 RR |
744 | if (IS_ERR(vcpu)) |
745 | return PTR_ERR(vcpu); | |
c5ea7660 | 746 | |
15ad7146 AK |
747 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
748 | ||
26e5215f AK |
749 | r = kvm_arch_vcpu_setup(vcpu); |
750 | if (r) | |
751 | goto vcpu_destroy; | |
752 | ||
11ec2804 | 753 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
754 | if (kvm->vcpus[n]) { |
755 | r = -EEXIST; | |
11ec2804 | 756 | mutex_unlock(&kvm->lock); |
e9b11c17 | 757 | goto vcpu_destroy; |
fb3f0f51 RR |
758 | } |
759 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 760 | mutex_unlock(&kvm->lock); |
c5ea7660 | 761 | |
fb3f0f51 | 762 | /* Now it's all set up, let userspace reach it */ |
bccf2150 AK |
763 | r = create_vcpu_fd(vcpu); |
764 | if (r < 0) | |
fb3f0f51 RR |
765 | goto unlink; |
766 | return r; | |
39c3b86e | 767 | |
fb3f0f51 | 768 | unlink: |
11ec2804 | 769 | mutex_lock(&kvm->lock); |
fb3f0f51 | 770 | kvm->vcpus[n] = NULL; |
11ec2804 | 771 | mutex_unlock(&kvm->lock); |
e9b11c17 | 772 | vcpu_destroy: |
d40ccc62 | 773 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
774 | return r; |
775 | } | |
776 | ||
1961d276 AK |
777 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
778 | { | |
779 | if (sigset) { | |
780 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
781 | vcpu->sigset_active = 1; | |
782 | vcpu->sigset = *sigset; | |
783 | } else | |
784 | vcpu->sigset_active = 0; | |
785 | return 0; | |
786 | } | |
787 | ||
bccf2150 AK |
788 | static long kvm_vcpu_ioctl(struct file *filp, |
789 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 790 | { |
bccf2150 | 791 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 792 | void __user *argp = (void __user *)arg; |
313a3dc7 | 793 | int r; |
6aa8b732 | 794 | |
6d4e4c4f AK |
795 | if (vcpu->kvm->mm != current->mm) |
796 | return -EIO; | |
6aa8b732 | 797 | switch (ioctl) { |
9a2bb7f4 | 798 | case KVM_RUN: |
f0fe5108 AK |
799 | r = -EINVAL; |
800 | if (arg) | |
801 | goto out; | |
b6c7a5dc | 802 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 803 | break; |
6aa8b732 AK |
804 | case KVM_GET_REGS: { |
805 | struct kvm_regs kvm_regs; | |
806 | ||
bccf2150 | 807 | memset(&kvm_regs, 0, sizeof kvm_regs); |
b6c7a5dc | 808 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, &kvm_regs); |
6aa8b732 AK |
809 | if (r) |
810 | goto out; | |
811 | r = -EFAULT; | |
2f366987 | 812 | if (copy_to_user(argp, &kvm_regs, sizeof kvm_regs)) |
6aa8b732 AK |
813 | goto out; |
814 | r = 0; | |
815 | break; | |
816 | } | |
817 | case KVM_SET_REGS: { | |
818 | struct kvm_regs kvm_regs; | |
819 | ||
820 | r = -EFAULT; | |
2f366987 | 821 | if (copy_from_user(&kvm_regs, argp, sizeof kvm_regs)) |
6aa8b732 | 822 | goto out; |
b6c7a5dc | 823 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, &kvm_regs); |
6aa8b732 AK |
824 | if (r) |
825 | goto out; | |
826 | r = 0; | |
827 | break; | |
828 | } | |
829 | case KVM_GET_SREGS: { | |
830 | struct kvm_sregs kvm_sregs; | |
831 | ||
bccf2150 | 832 | memset(&kvm_sregs, 0, sizeof kvm_sregs); |
b6c7a5dc | 833 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
834 | if (r) |
835 | goto out; | |
836 | r = -EFAULT; | |
2f366987 | 837 | if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs)) |
6aa8b732 AK |
838 | goto out; |
839 | r = 0; | |
840 | break; | |
841 | } | |
842 | case KVM_SET_SREGS: { | |
843 | struct kvm_sregs kvm_sregs; | |
844 | ||
845 | r = -EFAULT; | |
2f366987 | 846 | if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs)) |
6aa8b732 | 847 | goto out; |
b6c7a5dc | 848 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
849 | if (r) |
850 | goto out; | |
851 | r = 0; | |
852 | break; | |
853 | } | |
854 | case KVM_TRANSLATE: { | |
855 | struct kvm_translation tr; | |
856 | ||
857 | r = -EFAULT; | |
2f366987 | 858 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 859 | goto out; |
8b006791 | 860 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
861 | if (r) |
862 | goto out; | |
863 | r = -EFAULT; | |
2f366987 | 864 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
865 | goto out; |
866 | r = 0; | |
867 | break; | |
868 | } | |
6aa8b732 AK |
869 | case KVM_DEBUG_GUEST: { |
870 | struct kvm_debug_guest dbg; | |
871 | ||
872 | r = -EFAULT; | |
2f366987 | 873 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 874 | goto out; |
b6c7a5dc | 875 | r = kvm_arch_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
876 | if (r) |
877 | goto out; | |
878 | r = 0; | |
879 | break; | |
880 | } | |
1961d276 AK |
881 | case KVM_SET_SIGNAL_MASK: { |
882 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
883 | struct kvm_signal_mask kvm_sigmask; | |
884 | sigset_t sigset, *p; | |
885 | ||
886 | p = NULL; | |
887 | if (argp) { | |
888 | r = -EFAULT; | |
889 | if (copy_from_user(&kvm_sigmask, argp, | |
890 | sizeof kvm_sigmask)) | |
891 | goto out; | |
892 | r = -EINVAL; | |
893 | if (kvm_sigmask.len != sizeof sigset) | |
894 | goto out; | |
895 | r = -EFAULT; | |
896 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
897 | sizeof sigset)) | |
898 | goto out; | |
899 | p = &sigset; | |
900 | } | |
901 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
902 | break; | |
903 | } | |
b8836737 AK |
904 | case KVM_GET_FPU: { |
905 | struct kvm_fpu fpu; | |
906 | ||
907 | memset(&fpu, 0, sizeof fpu); | |
d0752060 | 908 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, &fpu); |
b8836737 AK |
909 | if (r) |
910 | goto out; | |
911 | r = -EFAULT; | |
912 | if (copy_to_user(argp, &fpu, sizeof fpu)) | |
913 | goto out; | |
914 | r = 0; | |
915 | break; | |
916 | } | |
917 | case KVM_SET_FPU: { | |
918 | struct kvm_fpu fpu; | |
919 | ||
920 | r = -EFAULT; | |
921 | if (copy_from_user(&fpu, argp, sizeof fpu)) | |
922 | goto out; | |
d0752060 | 923 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, &fpu); |
b8836737 AK |
924 | if (r) |
925 | goto out; | |
926 | r = 0; | |
927 | break; | |
928 | } | |
bccf2150 | 929 | default: |
313a3dc7 | 930 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
931 | } |
932 | out: | |
933 | return r; | |
934 | } | |
935 | ||
936 | static long kvm_vm_ioctl(struct file *filp, | |
937 | unsigned int ioctl, unsigned long arg) | |
938 | { | |
939 | struct kvm *kvm = filp->private_data; | |
940 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 941 | int r; |
bccf2150 | 942 | |
6d4e4c4f AK |
943 | if (kvm->mm != current->mm) |
944 | return -EIO; | |
bccf2150 AK |
945 | switch (ioctl) { |
946 | case KVM_CREATE_VCPU: | |
947 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
948 | if (r < 0) | |
949 | goto out; | |
950 | break; | |
6fc138d2 IE |
951 | case KVM_SET_USER_MEMORY_REGION: { |
952 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
953 | ||
954 | r = -EFAULT; | |
955 | if (copy_from_user(&kvm_userspace_mem, argp, | |
956 | sizeof kvm_userspace_mem)) | |
957 | goto out; | |
958 | ||
959 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
960 | if (r) |
961 | goto out; | |
962 | break; | |
963 | } | |
964 | case KVM_GET_DIRTY_LOG: { | |
965 | struct kvm_dirty_log log; | |
966 | ||
967 | r = -EFAULT; | |
2f366987 | 968 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 969 | goto out; |
2c6f5df9 | 970 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
971 | if (r) |
972 | goto out; | |
973 | break; | |
974 | } | |
f17abe9a | 975 | default: |
1fe779f8 | 976 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
977 | } |
978 | out: | |
979 | return r; | |
980 | } | |
981 | ||
e4a533a4 | 982 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a AK |
983 | { |
984 | struct kvm *kvm = vma->vm_file->private_data; | |
f17abe9a AK |
985 | struct page *page; |
986 | ||
e4a533a4 | 987 | if (!kvm_is_visible_gfn(kvm, vmf->pgoff)) |
988 | return VM_FAULT_SIGBUS; | |
10589a46 | 989 | page = gfn_to_page(kvm, vmf->pgoff); |
8a7ae055 | 990 | if (is_error_page(page)) { |
b4231d61 | 991 | kvm_release_page_clean(page); |
e4a533a4 | 992 | return VM_FAULT_SIGBUS; |
8a7ae055 | 993 | } |
e4a533a4 | 994 | vmf->page = page; |
995 | return 0; | |
f17abe9a AK |
996 | } |
997 | ||
998 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
e4a533a4 | 999 | .fault = kvm_vm_fault, |
f17abe9a AK |
1000 | }; |
1001 | ||
1002 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1003 | { | |
1004 | vma->vm_ops = &kvm_vm_vm_ops; | |
1005 | return 0; | |
1006 | } | |
1007 | ||
1008 | static struct file_operations kvm_vm_fops = { | |
1009 | .release = kvm_vm_release, | |
1010 | .unlocked_ioctl = kvm_vm_ioctl, | |
1011 | .compat_ioctl = kvm_vm_ioctl, | |
1012 | .mmap = kvm_vm_mmap, | |
1013 | }; | |
1014 | ||
1015 | static int kvm_dev_ioctl_create_vm(void) | |
1016 | { | |
1017 | int fd, r; | |
1018 | struct inode *inode; | |
1019 | struct file *file; | |
1020 | struct kvm *kvm; | |
1021 | ||
f17abe9a | 1022 | kvm = kvm_create_vm(); |
d6d28168 AK |
1023 | if (IS_ERR(kvm)) |
1024 | return PTR_ERR(kvm); | |
1025 | r = anon_inode_getfd(&fd, &inode, &file, "kvm-vm", &kvm_vm_fops, kvm); | |
1026 | if (r) { | |
1027 | kvm_destroy_vm(kvm); | |
1028 | return r; | |
f17abe9a AK |
1029 | } |
1030 | ||
bccf2150 | 1031 | kvm->filp = file; |
f17abe9a | 1032 | |
f17abe9a | 1033 | return fd; |
f17abe9a AK |
1034 | } |
1035 | ||
1036 | static long kvm_dev_ioctl(struct file *filp, | |
1037 | unsigned int ioctl, unsigned long arg) | |
1038 | { | |
1039 | void __user *argp = (void __user *)arg; | |
07c45a36 | 1040 | long r = -EINVAL; |
f17abe9a AK |
1041 | |
1042 | switch (ioctl) { | |
1043 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1044 | r = -EINVAL; |
1045 | if (arg) | |
1046 | goto out; | |
f17abe9a AK |
1047 | r = KVM_API_VERSION; |
1048 | break; | |
1049 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1050 | r = -EINVAL; |
1051 | if (arg) | |
1052 | goto out; | |
f17abe9a AK |
1053 | r = kvm_dev_ioctl_create_vm(); |
1054 | break; | |
018d00d2 ZX |
1055 | case KVM_CHECK_EXTENSION: |
1056 | r = kvm_dev_ioctl_check_extension((long)argp); | |
5d308f45 | 1057 | break; |
07c45a36 AK |
1058 | case KVM_GET_VCPU_MMAP_SIZE: |
1059 | r = -EINVAL; | |
1060 | if (arg) | |
1061 | goto out; | |
039576c0 | 1062 | r = 2 * PAGE_SIZE; |
07c45a36 | 1063 | break; |
6aa8b732 | 1064 | default: |
043405e1 | 1065 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1066 | } |
1067 | out: | |
1068 | return r; | |
1069 | } | |
1070 | ||
6aa8b732 | 1071 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1072 | .unlocked_ioctl = kvm_dev_ioctl, |
1073 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1074 | }; |
1075 | ||
1076 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1077 | KVM_MINOR, |
6aa8b732 AK |
1078 | "kvm", |
1079 | &kvm_chardev_ops, | |
1080 | }; | |
1081 | ||
1b6c0168 AK |
1082 | static void hardware_enable(void *junk) |
1083 | { | |
1084 | int cpu = raw_smp_processor_id(); | |
1085 | ||
1086 | if (cpu_isset(cpu, cpus_hardware_enabled)) | |
1087 | return; | |
1088 | cpu_set(cpu, cpus_hardware_enabled); | |
e9b11c17 | 1089 | kvm_arch_hardware_enable(NULL); |
1b6c0168 AK |
1090 | } |
1091 | ||
1092 | static void hardware_disable(void *junk) | |
1093 | { | |
1094 | int cpu = raw_smp_processor_id(); | |
1095 | ||
1096 | if (!cpu_isset(cpu, cpus_hardware_enabled)) | |
1097 | return; | |
1098 | cpu_clear(cpu, cpus_hardware_enabled); | |
1099 | decache_vcpus_on_cpu(cpu); | |
e9b11c17 | 1100 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1101 | } |
1102 | ||
774c47f1 AK |
1103 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1104 | void *v) | |
1105 | { | |
1106 | int cpu = (long)v; | |
1107 | ||
1a6f4d7f | 1108 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1109 | switch (val) { |
cec9ad27 | 1110 | case CPU_DYING: |
6ec8a856 AK |
1111 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1112 | cpu); | |
1113 | hardware_disable(NULL); | |
1114 | break; | |
774c47f1 | 1115 | case CPU_UP_CANCELED: |
43934a38 JK |
1116 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1117 | cpu); | |
1b6c0168 | 1118 | smp_call_function_single(cpu, hardware_disable, NULL, 0, 1); |
774c47f1 | 1119 | break; |
43934a38 JK |
1120 | case CPU_ONLINE: |
1121 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1122 | cpu); | |
1b6c0168 | 1123 | smp_call_function_single(cpu, hardware_enable, NULL, 0, 1); |
774c47f1 AK |
1124 | break; |
1125 | } | |
1126 | return NOTIFY_OK; | |
1127 | } | |
1128 | ||
9a2b85c6 | 1129 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1130 | void *v) |
9a2b85c6 RR |
1131 | { |
1132 | if (val == SYS_RESTART) { | |
1133 | /* | |
1134 | * Some (well, at least mine) BIOSes hang on reboot if | |
1135 | * in vmx root mode. | |
1136 | */ | |
1137 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1138 | on_each_cpu(hardware_disable, NULL, 0, 1); | |
1139 | } | |
1140 | return NOTIFY_OK; | |
1141 | } | |
1142 | ||
1143 | static struct notifier_block kvm_reboot_notifier = { | |
1144 | .notifier_call = kvm_reboot, | |
1145 | .priority = 0, | |
1146 | }; | |
1147 | ||
2eeb2e94 GH |
1148 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
1149 | { | |
1150 | memset(bus, 0, sizeof(*bus)); | |
1151 | } | |
1152 | ||
1153 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
1154 | { | |
1155 | int i; | |
1156 | ||
1157 | for (i = 0; i < bus->dev_count; i++) { | |
1158 | struct kvm_io_device *pos = bus->devs[i]; | |
1159 | ||
1160 | kvm_iodevice_destructor(pos); | |
1161 | } | |
1162 | } | |
1163 | ||
1164 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr) | |
1165 | { | |
1166 | int i; | |
1167 | ||
1168 | for (i = 0; i < bus->dev_count; i++) { | |
1169 | struct kvm_io_device *pos = bus->devs[i]; | |
1170 | ||
1171 | if (pos->in_range(pos, addr)) | |
1172 | return pos; | |
1173 | } | |
1174 | ||
1175 | return NULL; | |
1176 | } | |
1177 | ||
1178 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
1179 | { | |
1180 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
1181 | ||
1182 | bus->devs[bus->dev_count++] = dev; | |
1183 | } | |
1184 | ||
774c47f1 AK |
1185 | static struct notifier_block kvm_cpu_notifier = { |
1186 | .notifier_call = kvm_cpu_hotplug, | |
1187 | .priority = 20, /* must be > scheduler priority */ | |
1188 | }; | |
1189 | ||
8b88b099 | 1190 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
1191 | { |
1192 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
1193 | struct kvm *kvm; |
1194 | ||
8b88b099 | 1195 | *val = 0; |
ba1389b7 AK |
1196 | spin_lock(&kvm_lock); |
1197 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 1198 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 1199 | spin_unlock(&kvm_lock); |
8b88b099 | 1200 | return 0; |
ba1389b7 AK |
1201 | } |
1202 | ||
1203 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
1204 | ||
8b88b099 | 1205 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
1206 | { |
1207 | unsigned offset = (long)_offset; | |
1165f5fe AK |
1208 | struct kvm *kvm; |
1209 | struct kvm_vcpu *vcpu; | |
1210 | int i; | |
1211 | ||
8b88b099 | 1212 | *val = 0; |
1165f5fe AK |
1213 | spin_lock(&kvm_lock); |
1214 | list_for_each_entry(kvm, &vm_list, vm_list) | |
1215 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
1216 | vcpu = kvm->vcpus[i]; |
1217 | if (vcpu) | |
8b88b099 | 1218 | *val += *(u32 *)((void *)vcpu + offset); |
1165f5fe AK |
1219 | } |
1220 | spin_unlock(&kvm_lock); | |
8b88b099 | 1221 | return 0; |
1165f5fe AK |
1222 | } |
1223 | ||
ba1389b7 AK |
1224 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
1225 | ||
1226 | static struct file_operations *stat_fops[] = { | |
1227 | [KVM_STAT_VCPU] = &vcpu_stat_fops, | |
1228 | [KVM_STAT_VM] = &vm_stat_fops, | |
1229 | }; | |
1165f5fe | 1230 | |
a16b043c | 1231 | static void kvm_init_debug(void) |
6aa8b732 AK |
1232 | { |
1233 | struct kvm_stats_debugfs_item *p; | |
1234 | ||
8b6d44c7 | 1235 | debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 1236 | for (p = debugfs_entries; p->name; ++p) |
1165f5fe AK |
1237 | p->dentry = debugfs_create_file(p->name, 0444, debugfs_dir, |
1238 | (void *)(long)p->offset, | |
ba1389b7 | 1239 | stat_fops[p->kind]); |
6aa8b732 AK |
1240 | } |
1241 | ||
1242 | static void kvm_exit_debug(void) | |
1243 | { | |
1244 | struct kvm_stats_debugfs_item *p; | |
1245 | ||
1246 | for (p = debugfs_entries; p->name; ++p) | |
1247 | debugfs_remove(p->dentry); | |
1248 | debugfs_remove(debugfs_dir); | |
1249 | } | |
1250 | ||
59ae6c6b AK |
1251 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
1252 | { | |
4267c41a | 1253 | hardware_disable(NULL); |
59ae6c6b AK |
1254 | return 0; |
1255 | } | |
1256 | ||
1257 | static int kvm_resume(struct sys_device *dev) | |
1258 | { | |
4267c41a | 1259 | hardware_enable(NULL); |
59ae6c6b AK |
1260 | return 0; |
1261 | } | |
1262 | ||
1263 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 1264 | .name = "kvm", |
59ae6c6b AK |
1265 | .suspend = kvm_suspend, |
1266 | .resume = kvm_resume, | |
1267 | }; | |
1268 | ||
1269 | static struct sys_device kvm_sysdev = { | |
1270 | .id = 0, | |
1271 | .cls = &kvm_sysdev_class, | |
1272 | }; | |
1273 | ||
cea7bb21 | 1274 | struct page *bad_page; |
6aa8b732 | 1275 | |
15ad7146 AK |
1276 | static inline |
1277 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
1278 | { | |
1279 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
1280 | } | |
1281 | ||
1282 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
1283 | { | |
1284 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1285 | ||
e9b11c17 | 1286 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
1287 | } |
1288 | ||
1289 | static void kvm_sched_out(struct preempt_notifier *pn, | |
1290 | struct task_struct *next) | |
1291 | { | |
1292 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
1293 | ||
e9b11c17 | 1294 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
1295 | } |
1296 | ||
f8c16bba | 1297 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 1298 | struct module *module) |
6aa8b732 AK |
1299 | { |
1300 | int r; | |
002c7f7c | 1301 | int cpu; |
6aa8b732 | 1302 | |
cb498ea2 ZX |
1303 | kvm_init_debug(); |
1304 | ||
f8c16bba ZX |
1305 | r = kvm_arch_init(opaque); |
1306 | if (r) | |
d2308784 | 1307 | goto out_fail; |
cb498ea2 ZX |
1308 | |
1309 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
1310 | ||
1311 | if (bad_page == NULL) { | |
1312 | r = -ENOMEM; | |
1313 | goto out; | |
1314 | } | |
1315 | ||
e9b11c17 | 1316 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 1317 | if (r < 0) |
d2308784 | 1318 | goto out_free_0; |
6aa8b732 | 1319 | |
002c7f7c YS |
1320 | for_each_online_cpu(cpu) { |
1321 | smp_call_function_single(cpu, | |
e9b11c17 | 1322 | kvm_arch_check_processor_compat, |
002c7f7c YS |
1323 | &r, 0, 1); |
1324 | if (r < 0) | |
d2308784 | 1325 | goto out_free_1; |
002c7f7c YS |
1326 | } |
1327 | ||
1b6c0168 | 1328 | on_each_cpu(hardware_enable, NULL, 0, 1); |
774c47f1 AK |
1329 | r = register_cpu_notifier(&kvm_cpu_notifier); |
1330 | if (r) | |
d2308784 | 1331 | goto out_free_2; |
6aa8b732 AK |
1332 | register_reboot_notifier(&kvm_reboot_notifier); |
1333 | ||
59ae6c6b AK |
1334 | r = sysdev_class_register(&kvm_sysdev_class); |
1335 | if (r) | |
d2308784 | 1336 | goto out_free_3; |
59ae6c6b AK |
1337 | |
1338 | r = sysdev_register(&kvm_sysdev); | |
1339 | if (r) | |
d2308784 | 1340 | goto out_free_4; |
59ae6c6b | 1341 | |
c16f862d RR |
1342 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
1343 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
1344 | __alignof__(struct kvm_vcpu), |
1345 | 0, NULL); | |
c16f862d RR |
1346 | if (!kvm_vcpu_cache) { |
1347 | r = -ENOMEM; | |
d2308784 | 1348 | goto out_free_5; |
c16f862d RR |
1349 | } |
1350 | ||
6aa8b732 AK |
1351 | kvm_chardev_ops.owner = module; |
1352 | ||
1353 | r = misc_register(&kvm_dev); | |
1354 | if (r) { | |
d77c26fc | 1355 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
1356 | goto out_free; |
1357 | } | |
1358 | ||
15ad7146 AK |
1359 | kvm_preempt_ops.sched_in = kvm_sched_in; |
1360 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
1361 | ||
c7addb90 | 1362 | return 0; |
6aa8b732 AK |
1363 | |
1364 | out_free: | |
c16f862d | 1365 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 1366 | out_free_5: |
59ae6c6b | 1367 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 1368 | out_free_4: |
59ae6c6b | 1369 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 1370 | out_free_3: |
6aa8b732 | 1371 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 1372 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 1373 | out_free_2: |
1b6c0168 | 1374 | on_each_cpu(hardware_disable, NULL, 0, 1); |
d2308784 | 1375 | out_free_1: |
e9b11c17 | 1376 | kvm_arch_hardware_unsetup(); |
d2308784 ZX |
1377 | out_free_0: |
1378 | __free_page(bad_page); | |
ca45aaae | 1379 | out: |
f8c16bba | 1380 | kvm_arch_exit(); |
cb498ea2 | 1381 | kvm_exit_debug(); |
d2308784 | 1382 | out_fail: |
6aa8b732 AK |
1383 | return r; |
1384 | } | |
cb498ea2 | 1385 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 1386 | |
cb498ea2 | 1387 | void kvm_exit(void) |
6aa8b732 AK |
1388 | { |
1389 | misc_deregister(&kvm_dev); | |
c16f862d | 1390 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
1391 | sysdev_unregister(&kvm_sysdev); |
1392 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 1393 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 1394 | unregister_cpu_notifier(&kvm_cpu_notifier); |
1b6c0168 | 1395 | on_each_cpu(hardware_disable, NULL, 0, 1); |
e9b11c17 | 1396 | kvm_arch_hardware_unsetup(); |
f8c16bba | 1397 | kvm_arch_exit(); |
6aa8b732 | 1398 | kvm_exit_debug(); |
cea7bb21 | 1399 | __free_page(bad_page); |
6aa8b732 | 1400 | } |
cb498ea2 | 1401 | EXPORT_SYMBOL_GPL(kvm_exit); |