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 | 24 | #include <linux/percpu.h> |
6aa8b732 AK |
25 | #include <linux/mm.h> |
26 | #include <linux/miscdevice.h> | |
27 | #include <linux/vmalloc.h> | |
6aa8b732 | 28 | #include <linux/reboot.h> |
6aa8b732 AK |
29 | #include <linux/debugfs.h> |
30 | #include <linux/highmem.h> | |
31 | #include <linux/file.h> | |
59ae6c6b | 32 | #include <linux/sysdev.h> |
774c47f1 | 33 | #include <linux/cpu.h> |
e8edc6e0 | 34 | #include <linux/sched.h> |
d9e368d6 AK |
35 | #include <linux/cpumask.h> |
36 | #include <linux/smp.h> | |
d6d28168 | 37 | #include <linux/anon_inodes.h> |
04d2cc77 | 38 | #include <linux/profile.h> |
7aa81cc0 | 39 | #include <linux/kvm_para.h> |
6fc138d2 | 40 | #include <linux/pagemap.h> |
8d4e1288 | 41 | #include <linux/mman.h> |
35149e21 | 42 | #include <linux/swap.h> |
e56d532f | 43 | #include <linux/bitops.h> |
547de29e | 44 | #include <linux/spinlock.h> |
6ff5894c | 45 | #include <linux/compat.h> |
bc6678a3 | 46 | #include <linux/srcu.h> |
8f0b1ab6 | 47 | #include <linux/hugetlb.h> |
5a0e3ad6 | 48 | #include <linux/slab.h> |
6aa8b732 | 49 | |
e495606d | 50 | #include <asm/processor.h> |
e495606d AK |
51 | #include <asm/io.h> |
52 | #include <asm/uaccess.h> | |
3e021bf5 | 53 | #include <asm/pgtable.h> |
c8240bd6 | 54 | #include <asm-generic/bitops/le.h> |
6aa8b732 | 55 | |
5f94c174 | 56 | #include "coalesced_mmio.h" |
5f94c174 | 57 | |
229456fc MT |
58 | #define CREATE_TRACE_POINTS |
59 | #include <trace/events/kvm.h> | |
60 | ||
6aa8b732 AK |
61 | MODULE_AUTHOR("Qumranet"); |
62 | MODULE_LICENSE("GPL"); | |
63 | ||
fa40a821 MT |
64 | /* |
65 | * Ordering of locks: | |
66 | * | |
fae3a353 | 67 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
68 | */ |
69 | ||
e9b11c17 ZX |
70 | DEFINE_SPINLOCK(kvm_lock); |
71 | LIST_HEAD(vm_list); | |
133de902 | 72 | |
7f59f492 | 73 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
74 | static int kvm_usage_count = 0; |
75 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 76 | |
c16f862d RR |
77 | struct kmem_cache *kvm_vcpu_cache; |
78 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 79 | |
15ad7146 AK |
80 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
81 | ||
76f7c879 | 82 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 83 | |
bccf2150 AK |
84 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
85 | unsigned long arg); | |
10474ae8 AG |
86 | static int hardware_enable_all(void); |
87 | static void hardware_disable_all(void); | |
bccf2150 | 88 | |
e93f8a0f MT |
89 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
90 | ||
e8ba5d31 | 91 | static bool kvm_rebooting; |
4ecac3fd | 92 | |
54dee993 MT |
93 | static bool largepages_enabled = true; |
94 | ||
c77fb9dc | 95 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 96 | { |
fc5659c8 JR |
97 | if (pfn_valid(pfn)) { |
98 | struct page *page = compound_head(pfn_to_page(pfn)); | |
99 | return PageReserved(page); | |
100 | } | |
cbff90a7 BAY |
101 | |
102 | return true; | |
103 | } | |
104 | ||
bccf2150 AK |
105 | /* |
106 | * Switches to specified vcpu, until a matching vcpu_put() | |
107 | */ | |
313a3dc7 | 108 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 109 | { |
15ad7146 AK |
110 | int cpu; |
111 | ||
bccf2150 | 112 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
113 | cpu = get_cpu(); |
114 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 115 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 116 | put_cpu(); |
6aa8b732 AK |
117 | } |
118 | ||
313a3dc7 | 119 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 120 | { |
15ad7146 | 121 | preempt_disable(); |
313a3dc7 | 122 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
123 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
124 | preempt_enable(); | |
6aa8b732 AK |
125 | mutex_unlock(&vcpu->mutex); |
126 | } | |
127 | ||
d9e368d6 AK |
128 | static void ack_flush(void *_completed) |
129 | { | |
d9e368d6 AK |
130 | } |
131 | ||
49846896 | 132 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 133 | { |
597a5f55 | 134 | int i, cpu, me; |
6ef7a1bc RR |
135 | cpumask_var_t cpus; |
136 | bool called = true; | |
d9e368d6 | 137 | struct kvm_vcpu *vcpu; |
d9e368d6 | 138 | |
79f55997 | 139 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 140 | |
70e335e1 | 141 | raw_spin_lock(&kvm->requests_lock); |
e601e3be | 142 | me = smp_processor_id(); |
988a2cae | 143 | kvm_for_each_vcpu(i, vcpu, kvm) { |
49846896 | 144 | if (test_and_set_bit(req, &vcpu->requests)) |
d9e368d6 AK |
145 | continue; |
146 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
147 | if (cpus != NULL && cpu != -1 && cpu != me) |
148 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 149 | } |
6ef7a1bc RR |
150 | if (unlikely(cpus == NULL)) |
151 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
152 | else if (!cpumask_empty(cpus)) | |
153 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
154 | else | |
155 | called = false; | |
70e335e1 | 156 | raw_spin_unlock(&kvm->requests_lock); |
6ef7a1bc | 157 | free_cpumask_var(cpus); |
49846896 | 158 | return called; |
d9e368d6 AK |
159 | } |
160 | ||
49846896 | 161 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 162 | { |
49846896 RR |
163 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
164 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
165 | } |
166 | ||
49846896 RR |
167 | void kvm_reload_remote_mmus(struct kvm *kvm) |
168 | { | |
169 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
170 | } | |
2e53d63a | 171 | |
fb3f0f51 RR |
172 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
173 | { | |
174 | struct page *page; | |
175 | int r; | |
176 | ||
177 | mutex_init(&vcpu->mutex); | |
178 | vcpu->cpu = -1; | |
fb3f0f51 RR |
179 | vcpu->kvm = kvm; |
180 | vcpu->vcpu_id = id; | |
b6958ce4 | 181 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
182 | |
183 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
184 | if (!page) { | |
185 | r = -ENOMEM; | |
186 | goto fail; | |
187 | } | |
188 | vcpu->run = page_address(page); | |
189 | ||
e9b11c17 | 190 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 191 | if (r < 0) |
e9b11c17 | 192 | goto fail_free_run; |
fb3f0f51 RR |
193 | return 0; |
194 | ||
fb3f0f51 RR |
195 | fail_free_run: |
196 | free_page((unsigned long)vcpu->run); | |
197 | fail: | |
76fafa5e | 198 | return r; |
fb3f0f51 RR |
199 | } |
200 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
201 | ||
202 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
203 | { | |
e9b11c17 | 204 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
205 | free_page((unsigned long)vcpu->run); |
206 | } | |
207 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
208 | ||
e930bffe AA |
209 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
210 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
211 | { | |
212 | return container_of(mn, struct kvm, mmu_notifier); | |
213 | } | |
214 | ||
215 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
216 | struct mm_struct *mm, | |
217 | unsigned long address) | |
218 | { | |
219 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 220 | int need_tlb_flush, idx; |
e930bffe AA |
221 | |
222 | /* | |
223 | * When ->invalidate_page runs, the linux pte has been zapped | |
224 | * already but the page is still allocated until | |
225 | * ->invalidate_page returns. So if we increase the sequence | |
226 | * here the kvm page fault will notice if the spte can't be | |
227 | * established because the page is going to be freed. If | |
228 | * instead the kvm page fault establishes the spte before | |
229 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
230 | * before returning. | |
231 | * | |
232 | * The sequence increase only need to be seen at spin_unlock | |
233 | * time, and not at spin_lock time. | |
234 | * | |
235 | * Increasing the sequence after the spin_unlock would be | |
236 | * unsafe because the kvm page fault could then establish the | |
237 | * pte after kvm_unmap_hva returned, without noticing the page | |
238 | * is going to be freed. | |
239 | */ | |
bc6678a3 | 240 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
241 | spin_lock(&kvm->mmu_lock); |
242 | kvm->mmu_notifier_seq++; | |
243 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
244 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 245 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
246 | |
247 | /* we've to flush the tlb before the pages can be freed */ | |
248 | if (need_tlb_flush) | |
249 | kvm_flush_remote_tlbs(kvm); | |
250 | ||
251 | } | |
252 | ||
3da0dd43 IE |
253 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
254 | struct mm_struct *mm, | |
255 | unsigned long address, | |
256 | pte_t pte) | |
257 | { | |
258 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 259 | int idx; |
3da0dd43 | 260 | |
bc6678a3 | 261 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
262 | spin_lock(&kvm->mmu_lock); |
263 | kvm->mmu_notifier_seq++; | |
264 | kvm_set_spte_hva(kvm, address, pte); | |
265 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 266 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
267 | } |
268 | ||
e930bffe AA |
269 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
270 | struct mm_struct *mm, | |
271 | unsigned long start, | |
272 | unsigned long end) | |
273 | { | |
274 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 275 | int need_tlb_flush = 0, idx; |
e930bffe | 276 | |
bc6678a3 | 277 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
278 | spin_lock(&kvm->mmu_lock); |
279 | /* | |
280 | * The count increase must become visible at unlock time as no | |
281 | * spte can be established without taking the mmu_lock and | |
282 | * count is also read inside the mmu_lock critical section. | |
283 | */ | |
284 | kvm->mmu_notifier_count++; | |
285 | for (; start < end; start += PAGE_SIZE) | |
286 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
287 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 288 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
289 | |
290 | /* we've to flush the tlb before the pages can be freed */ | |
291 | if (need_tlb_flush) | |
292 | kvm_flush_remote_tlbs(kvm); | |
293 | } | |
294 | ||
295 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
296 | struct mm_struct *mm, | |
297 | unsigned long start, | |
298 | unsigned long end) | |
299 | { | |
300 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
301 | ||
302 | spin_lock(&kvm->mmu_lock); | |
303 | /* | |
304 | * This sequence increase will notify the kvm page fault that | |
305 | * the page that is going to be mapped in the spte could have | |
306 | * been freed. | |
307 | */ | |
308 | kvm->mmu_notifier_seq++; | |
309 | /* | |
310 | * The above sequence increase must be visible before the | |
311 | * below count decrease but both values are read by the kvm | |
312 | * page fault under mmu_lock spinlock so we don't need to add | |
313 | * a smb_wmb() here in between the two. | |
314 | */ | |
315 | kvm->mmu_notifier_count--; | |
316 | spin_unlock(&kvm->mmu_lock); | |
317 | ||
318 | BUG_ON(kvm->mmu_notifier_count < 0); | |
319 | } | |
320 | ||
321 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
322 | struct mm_struct *mm, | |
323 | unsigned long address) | |
324 | { | |
325 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 326 | int young, idx; |
e930bffe | 327 | |
bc6678a3 | 328 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
329 | spin_lock(&kvm->mmu_lock); |
330 | young = kvm_age_hva(kvm, address); | |
331 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 332 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
333 | |
334 | if (young) | |
335 | kvm_flush_remote_tlbs(kvm); | |
336 | ||
337 | return young; | |
338 | } | |
339 | ||
85db06e5 MT |
340 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
341 | struct mm_struct *mm) | |
342 | { | |
343 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
344 | kvm_arch_flush_shadow(kvm); | |
345 | } | |
346 | ||
e930bffe AA |
347 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
348 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
349 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
350 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
351 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
3da0dd43 | 352 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 353 | .release = kvm_mmu_notifier_release, |
e930bffe | 354 | }; |
4c07b0a4 AK |
355 | |
356 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
357 | { | |
358 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
359 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
360 | } | |
361 | ||
362 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
363 | ||
364 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
365 | { | |
366 | return 0; | |
367 | } | |
368 | ||
e930bffe AA |
369 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
370 | ||
f17abe9a | 371 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 372 | { |
e93f8a0f | 373 | int r = 0, i; |
d19a9cd2 | 374 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 375 | |
d19a9cd2 ZX |
376 | if (IS_ERR(kvm)) |
377 | goto out; | |
10474ae8 AG |
378 | |
379 | r = hardware_enable_all(); | |
380 | if (r) | |
381 | goto out_err_nodisable; | |
382 | ||
75858a84 AK |
383 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
384 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 385 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 386 | #endif |
6aa8b732 | 387 | |
46a26bf5 MT |
388 | r = -ENOMEM; |
389 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
390 | if (!kvm->memslots) | |
391 | goto out_err; | |
bc6678a3 MT |
392 | if (init_srcu_struct(&kvm->srcu)) |
393 | goto out_err; | |
e93f8a0f MT |
394 | for (i = 0; i < KVM_NR_BUSES; i++) { |
395 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
396 | GFP_KERNEL); | |
397 | if (!kvm->buses[i]) { | |
398 | cleanup_srcu_struct(&kvm->srcu); | |
399 | goto out_err; | |
400 | } | |
401 | } | |
46a26bf5 | 402 | |
4c07b0a4 | 403 | r = kvm_init_mmu_notifier(kvm); |
283d0c65 | 404 | if (r) { |
bc6678a3 | 405 | cleanup_srcu_struct(&kvm->srcu); |
283d0c65 | 406 | goto out_err; |
e930bffe | 407 | } |
e930bffe | 408 | |
6d4e4c4f AK |
409 | kvm->mm = current->mm; |
410 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 411 | spin_lock_init(&kvm->mmu_lock); |
70e335e1 | 412 | raw_spin_lock_init(&kvm->requests_lock); |
d34e6b17 | 413 | kvm_eventfd_init(kvm); |
11ec2804 | 414 | mutex_init(&kvm->lock); |
60eead79 | 415 | mutex_init(&kvm->irq_lock); |
79fac95e | 416 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 417 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
418 | spin_lock(&kvm_lock); |
419 | list_add(&kvm->vm_list, &vm_list); | |
420 | spin_unlock(&kvm_lock); | |
5f94c174 LV |
421 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
422 | kvm_coalesced_mmio_init(kvm); | |
423 | #endif | |
d19a9cd2 | 424 | out: |
f17abe9a | 425 | return kvm; |
10474ae8 AG |
426 | |
427 | out_err: | |
428 | hardware_disable_all(); | |
429 | out_err_nodisable: | |
e93f8a0f MT |
430 | for (i = 0; i < KVM_NR_BUSES; i++) |
431 | kfree(kvm->buses[i]); | |
46a26bf5 | 432 | kfree(kvm->memslots); |
10474ae8 AG |
433 | kfree(kvm); |
434 | return ERR_PTR(r); | |
f17abe9a AK |
435 | } |
436 | ||
6aa8b732 AK |
437 | /* |
438 | * Free any memory in @free but not in @dont. | |
439 | */ | |
440 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
441 | struct kvm_memory_slot *dont) | |
442 | { | |
ec04b260 JR |
443 | int i; |
444 | ||
290fc38d IE |
445 | if (!dont || free->rmap != dont->rmap) |
446 | vfree(free->rmap); | |
6aa8b732 AK |
447 | |
448 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
449 | vfree(free->dirty_bitmap); | |
450 | ||
ec04b260 JR |
451 | |
452 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
453 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
454 | vfree(free->lpage_info[i]); | |
455 | free->lpage_info[i] = NULL; | |
456 | } | |
457 | } | |
05da4558 | 458 | |
6aa8b732 | 459 | free->npages = 0; |
8b6d44c7 | 460 | free->dirty_bitmap = NULL; |
8d4e1288 | 461 | free->rmap = NULL; |
6aa8b732 AK |
462 | } |
463 | ||
d19a9cd2 | 464 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
465 | { |
466 | int i; | |
46a26bf5 MT |
467 | struct kvm_memslots *slots = kvm->memslots; |
468 | ||
469 | for (i = 0; i < slots->nmemslots; ++i) | |
470 | kvm_free_physmem_slot(&slots->memslots[i], NULL); | |
6aa8b732 | 471 | |
46a26bf5 | 472 | kfree(kvm->memslots); |
6aa8b732 AK |
473 | } |
474 | ||
f17abe9a AK |
475 | static void kvm_destroy_vm(struct kvm *kvm) |
476 | { | |
e93f8a0f | 477 | int i; |
6d4e4c4f AK |
478 | struct mm_struct *mm = kvm->mm; |
479 | ||
ad8ba2cd | 480 | kvm_arch_sync_events(kvm); |
133de902 AK |
481 | spin_lock(&kvm_lock); |
482 | list_del(&kvm->vm_list); | |
483 | spin_unlock(&kvm_lock); | |
399ec807 | 484 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
485 | for (i = 0; i < KVM_NR_BUSES; i++) |
486 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 487 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
488 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
489 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
490 | #else |
491 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 492 | #endif |
d19a9cd2 | 493 | kvm_arch_destroy_vm(kvm); |
10474ae8 | 494 | hardware_disable_all(); |
6d4e4c4f | 495 | mmdrop(mm); |
f17abe9a AK |
496 | } |
497 | ||
d39f13b0 IE |
498 | void kvm_get_kvm(struct kvm *kvm) |
499 | { | |
500 | atomic_inc(&kvm->users_count); | |
501 | } | |
502 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
503 | ||
504 | void kvm_put_kvm(struct kvm *kvm) | |
505 | { | |
506 | if (atomic_dec_and_test(&kvm->users_count)) | |
507 | kvm_destroy_vm(kvm); | |
508 | } | |
509 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
510 | ||
511 | ||
f17abe9a AK |
512 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
513 | { | |
514 | struct kvm *kvm = filp->private_data; | |
515 | ||
721eecbf GH |
516 | kvm_irqfd_release(kvm); |
517 | ||
d39f13b0 | 518 | kvm_put_kvm(kvm); |
6aa8b732 AK |
519 | return 0; |
520 | } | |
521 | ||
6aa8b732 AK |
522 | /* |
523 | * Allocate some memory and give it an address in the guest physical address | |
524 | * space. | |
525 | * | |
526 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 527 | * |
10589a46 | 528 | * Must be called holding mmap_sem for write. |
6aa8b732 | 529 | */ |
f78e0e2e SY |
530 | int __kvm_set_memory_region(struct kvm *kvm, |
531 | struct kvm_userspace_memory_region *mem, | |
532 | int user_alloc) | |
6aa8b732 | 533 | { |
bc6678a3 | 534 | int r, flush_shadow = 0; |
6aa8b732 | 535 | gfn_t base_gfn; |
28bcb112 HC |
536 | unsigned long npages; |
537 | unsigned long i; | |
6aa8b732 AK |
538 | struct kvm_memory_slot *memslot; |
539 | struct kvm_memory_slot old, new; | |
bc6678a3 | 540 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
541 | |
542 | r = -EINVAL; | |
543 | /* General sanity checks */ | |
544 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
545 | goto out; | |
546 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
547 | goto out; | |
e7cacd40 | 548 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 549 | goto out; |
e0d62c7f | 550 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
551 | goto out; |
552 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
553 | goto out; | |
554 | ||
46a26bf5 | 555 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
556 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
557 | npages = mem->memory_size >> PAGE_SHIFT; | |
558 | ||
559 | if (!npages) | |
560 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
561 | ||
6aa8b732 AK |
562 | new = old = *memslot; |
563 | ||
564 | new.base_gfn = base_gfn; | |
565 | new.npages = npages; | |
566 | new.flags = mem->flags; | |
567 | ||
568 | /* Disallow changing a memory slot's size. */ | |
569 | r = -EINVAL; | |
570 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 571 | goto out_free; |
6aa8b732 AK |
572 | |
573 | /* Check for overlaps */ | |
574 | r = -EEXIST; | |
575 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 576 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 577 | |
4cd481f6 | 578 | if (s == memslot || !s->npages) |
6aa8b732 AK |
579 | continue; |
580 | if (!((base_gfn + npages <= s->base_gfn) || | |
581 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 582 | goto out_free; |
6aa8b732 | 583 | } |
6aa8b732 | 584 | |
6aa8b732 AK |
585 | /* Free page dirty bitmap if unneeded */ |
586 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 587 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
588 | |
589 | r = -ENOMEM; | |
590 | ||
591 | /* Allocate if a slot is being created */ | |
eff0114a | 592 | #ifndef CONFIG_S390 |
8d4e1288 | 593 | if (npages && !new.rmap) { |
d77c26fc | 594 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
595 | |
596 | if (!new.rmap) | |
f78e0e2e | 597 | goto out_free; |
290fc38d | 598 | |
290fc38d | 599 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 600 | |
80b14b5b | 601 | new.user_alloc = user_alloc; |
bc6678a3 | 602 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 603 | } |
ec04b260 JR |
604 | if (!npages) |
605 | goto skip_lpage; | |
05da4558 | 606 | |
ec04b260 | 607 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
608 | unsigned long ugfn; |
609 | unsigned long j; | |
610 | int lpages; | |
ec04b260 | 611 | int level = i + 2; |
05da4558 | 612 | |
ec04b260 JR |
613 | /* Avoid unused variable warning if no large pages */ |
614 | (void)level; | |
615 | ||
616 | if (new.lpage_info[i]) | |
617 | continue; | |
618 | ||
619 | lpages = 1 + (base_gfn + npages - 1) / | |
620 | KVM_PAGES_PER_HPAGE(level); | |
621 | lpages -= base_gfn / KVM_PAGES_PER_HPAGE(level); | |
622 | ||
623 | new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i])); | |
624 | ||
625 | if (!new.lpage_info[i]) | |
05da4558 MT |
626 | goto out_free; |
627 | ||
ec04b260 JR |
628 | memset(new.lpage_info[i], 0, |
629 | lpages * sizeof(*new.lpage_info[i])); | |
05da4558 | 630 | |
ec04b260 JR |
631 | if (base_gfn % KVM_PAGES_PER_HPAGE(level)) |
632 | new.lpage_info[i][0].write_count = 1; | |
633 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE(level)) | |
634 | new.lpage_info[i][lpages - 1].write_count = 1; | |
ac04527f AK |
635 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
636 | /* | |
637 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
638 | * other, or if explicitly asked to, disable large page |
639 | * support for this slot | |
ac04527f | 640 | */ |
ec04b260 | 641 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 642 | !largepages_enabled) |
ec04b260 JR |
643 | for (j = 0; j < lpages; ++j) |
644 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 645 | } |
6aa8b732 | 646 | |
ec04b260 JR |
647 | skip_lpage: |
648 | ||
6aa8b732 AK |
649 | /* Allocate page dirty bitmap if needed */ |
650 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
651 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
652 | ||
653 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
654 | if (!new.dirty_bitmap) | |
f78e0e2e | 655 | goto out_free; |
6aa8b732 | 656 | memset(new.dirty_bitmap, 0, dirty_bytes); |
bc6678a3 | 657 | /* destroy any largepage mappings for dirty tracking */ |
e244584f | 658 | if (old.npages) |
bc6678a3 | 659 | flush_shadow = 1; |
6aa8b732 | 660 | } |
3eea8437 CB |
661 | #else /* not defined CONFIG_S390 */ |
662 | new.user_alloc = user_alloc; | |
663 | if (user_alloc) | |
664 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 665 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 666 | |
bc6678a3 MT |
667 | if (!npages) { |
668 | r = -ENOMEM; | |
669 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
670 | if (!slots) | |
671 | goto out_free; | |
672 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
673 | if (mem->slot >= slots->nmemslots) | |
674 | slots->nmemslots = mem->slot + 1; | |
675 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; | |
676 | ||
677 | old_memslots = kvm->memslots; | |
678 | rcu_assign_pointer(kvm->memslots, slots); | |
679 | synchronize_srcu_expedited(&kvm->srcu); | |
680 | /* From this point no new shadow pages pointing to a deleted | |
681 | * memslot will be created. | |
682 | * | |
683 | * validation of sp->gfn happens in: | |
684 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
685 | * - kvm_is_visible_gfn (mmu_check_roots) | |
686 | */ | |
34d4cb8f | 687 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
688 | kfree(old_memslots); |
689 | } | |
34d4cb8f | 690 | |
f7784b8e MT |
691 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
692 | if (r) | |
693 | goto out_free; | |
694 | ||
bc6678a3 MT |
695 | #ifdef CONFIG_DMAR |
696 | /* map the pages in iommu page table */ | |
697 | if (npages) { | |
698 | r = kvm_iommu_map_pages(kvm, &new); | |
699 | if (r) | |
700 | goto out_free; | |
701 | } | |
702 | #endif | |
604b38ac | 703 | |
bc6678a3 MT |
704 | r = -ENOMEM; |
705 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
706 | if (!slots) | |
707 | goto out_free; | |
708 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
709 | if (mem->slot >= slots->nmemslots) | |
710 | slots->nmemslots = mem->slot + 1; | |
711 | ||
712 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
713 | if (!npages) { | |
714 | new.rmap = NULL; | |
715 | new.dirty_bitmap = NULL; | |
716 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
717 | new.lpage_info[i] = NULL; | |
718 | } | |
719 | ||
720 | slots->memslots[mem->slot] = new; | |
721 | old_memslots = kvm->memslots; | |
722 | rcu_assign_pointer(kvm->memslots, slots); | |
723 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 724 | |
f7784b8e | 725 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 726 | |
bc6678a3 MT |
727 | kvm_free_physmem_slot(&old, &new); |
728 | kfree(old_memslots); | |
729 | ||
730 | if (flush_shadow) | |
731 | kvm_arch_flush_shadow(kvm); | |
732 | ||
6aa8b732 AK |
733 | return 0; |
734 | ||
f78e0e2e | 735 | out_free: |
6aa8b732 AK |
736 | kvm_free_physmem_slot(&new, &old); |
737 | out: | |
738 | return r; | |
210c7c4d IE |
739 | |
740 | } | |
f78e0e2e SY |
741 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
742 | ||
743 | int kvm_set_memory_region(struct kvm *kvm, | |
744 | struct kvm_userspace_memory_region *mem, | |
745 | int user_alloc) | |
746 | { | |
747 | int r; | |
748 | ||
79fac95e | 749 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 750 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 751 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
752 | return r; |
753 | } | |
210c7c4d IE |
754 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
755 | ||
1fe779f8 CO |
756 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
757 | struct | |
758 | kvm_userspace_memory_region *mem, | |
759 | int user_alloc) | |
210c7c4d | 760 | { |
e0d62c7f IE |
761 | if (mem->slot >= KVM_MEMORY_SLOTS) |
762 | return -EINVAL; | |
210c7c4d | 763 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
764 | } |
765 | ||
5bb064dc ZX |
766 | int kvm_get_dirty_log(struct kvm *kvm, |
767 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
768 | { |
769 | struct kvm_memory_slot *memslot; | |
770 | int r, i; | |
771 | int n; | |
772 | unsigned long any = 0; | |
773 | ||
6aa8b732 AK |
774 | r = -EINVAL; |
775 | if (log->slot >= KVM_MEMORY_SLOTS) | |
776 | goto out; | |
777 | ||
46a26bf5 | 778 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
779 | r = -ENOENT; |
780 | if (!memslot->dirty_bitmap) | |
781 | goto out; | |
782 | ||
cd1a4a98 | 783 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 784 | |
cd1a4a98 | 785 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
786 | any = memslot->dirty_bitmap[i]; |
787 | ||
788 | r = -EFAULT; | |
789 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
790 | goto out; | |
791 | ||
5bb064dc ZX |
792 | if (any) |
793 | *is_dirty = 1; | |
6aa8b732 AK |
794 | |
795 | r = 0; | |
6aa8b732 | 796 | out: |
6aa8b732 AK |
797 | return r; |
798 | } | |
799 | ||
54dee993 MT |
800 | void kvm_disable_largepages(void) |
801 | { | |
802 | largepages_enabled = false; | |
803 | } | |
804 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
805 | ||
cea7bb21 IE |
806 | int is_error_page(struct page *page) |
807 | { | |
808 | return page == bad_page; | |
809 | } | |
810 | EXPORT_SYMBOL_GPL(is_error_page); | |
811 | ||
35149e21 AL |
812 | int is_error_pfn(pfn_t pfn) |
813 | { | |
814 | return pfn == bad_pfn; | |
815 | } | |
816 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
817 | ||
f9d46eb0 IE |
818 | static inline unsigned long bad_hva(void) |
819 | { | |
820 | return PAGE_OFFSET; | |
821 | } | |
822 | ||
823 | int kvm_is_error_hva(unsigned long addr) | |
824 | { | |
825 | return addr == bad_hva(); | |
826 | } | |
827 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
828 | ||
2843099f | 829 | struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
830 | { |
831 | int i; | |
bc6678a3 | 832 | struct kvm_memslots *slots = rcu_dereference(kvm->memslots); |
6aa8b732 | 833 | |
46a26bf5 MT |
834 | for (i = 0; i < slots->nmemslots; ++i) { |
835 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
836 | |
837 | if (gfn >= memslot->base_gfn | |
838 | && gfn < memslot->base_gfn + memslot->npages) | |
839 | return memslot; | |
840 | } | |
8b6d44c7 | 841 | return NULL; |
6aa8b732 | 842 | } |
2843099f | 843 | EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased); |
e8207547 AK |
844 | |
845 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
846 | { | |
847 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 848 | return gfn_to_memslot_unaliased(kvm, gfn); |
e8207547 | 849 | } |
6aa8b732 | 850 | |
e0d62c7f IE |
851 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
852 | { | |
853 | int i; | |
bc6678a3 | 854 | struct kvm_memslots *slots = rcu_dereference(kvm->memslots); |
e0d62c7f | 855 | |
a983fb23 | 856 | gfn = unalias_gfn_instantiation(kvm, gfn); |
e0d62c7f | 857 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 858 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 859 | |
bc6678a3 MT |
860 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
861 | continue; | |
862 | ||
e0d62c7f IE |
863 | if (gfn >= memslot->base_gfn |
864 | && gfn < memslot->base_gfn + memslot->npages) | |
865 | return 1; | |
866 | } | |
867 | return 0; | |
868 | } | |
869 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
870 | ||
8f0b1ab6 JR |
871 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
872 | { | |
873 | struct vm_area_struct *vma; | |
874 | unsigned long addr, size; | |
875 | ||
876 | size = PAGE_SIZE; | |
877 | ||
878 | addr = gfn_to_hva(kvm, gfn); | |
879 | if (kvm_is_error_hva(addr)) | |
880 | return PAGE_SIZE; | |
881 | ||
882 | down_read(¤t->mm->mmap_sem); | |
883 | vma = find_vma(current->mm, addr); | |
884 | if (!vma) | |
885 | goto out; | |
886 | ||
887 | size = vma_kernel_pagesize(vma); | |
888 | ||
889 | out: | |
890 | up_read(¤t->mm->mmap_sem); | |
891 | ||
892 | return size; | |
893 | } | |
894 | ||
bc6678a3 MT |
895 | int memslot_id(struct kvm *kvm, gfn_t gfn) |
896 | { | |
897 | int i; | |
898 | struct kvm_memslots *slots = rcu_dereference(kvm->memslots); | |
899 | struct kvm_memory_slot *memslot = NULL; | |
900 | ||
901 | gfn = unalias_gfn(kvm, gfn); | |
902 | for (i = 0; i < slots->nmemslots; ++i) { | |
903 | memslot = &slots->memslots[i]; | |
904 | ||
905 | if (gfn >= memslot->base_gfn | |
906 | && gfn < memslot->base_gfn + memslot->npages) | |
907 | break; | |
908 | } | |
909 | ||
910 | return memslot - slots->memslots; | |
911 | } | |
912 | ||
05da4558 | 913 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
914 | { |
915 | struct kvm_memory_slot *slot; | |
916 | ||
a983fb23 | 917 | gfn = unalias_gfn_instantiation(kvm, gfn); |
2843099f | 918 | slot = gfn_to_memslot_unaliased(kvm, gfn); |
bc6678a3 | 919 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 IE |
920 | return bad_hva(); |
921 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
922 | } | |
0d150298 | 923 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 924 | |
506f0d6f | 925 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr) |
954bbbc2 | 926 | { |
8d4e1288 AL |
927 | struct page *page[1]; |
928 | int npages; | |
2e2e3738 | 929 | pfn_t pfn; |
954bbbc2 | 930 | |
60395224 AK |
931 | might_sleep(); |
932 | ||
4c2155ce | 933 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 934 | |
2e2e3738 AL |
935 | if (unlikely(npages != 1)) { |
936 | struct vm_area_struct *vma; | |
937 | ||
4c2155ce | 938 | down_read(¤t->mm->mmap_sem); |
2e2e3738 | 939 | vma = find_vma(current->mm, addr); |
4c2155ce | 940 | |
2e2e3738 AL |
941 | if (vma == NULL || addr < vma->vm_start || |
942 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 943 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
944 | get_page(bad_page); |
945 | return page_to_pfn(bad_page); | |
946 | } | |
947 | ||
948 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 949 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 950 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
951 | } else |
952 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 953 | |
2e2e3738 | 954 | return pfn; |
35149e21 AL |
955 | } |
956 | ||
506f0d6f MT |
957 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
958 | { | |
959 | unsigned long addr; | |
960 | ||
961 | addr = gfn_to_hva(kvm, gfn); | |
962 | if (kvm_is_error_hva(addr)) { | |
963 | get_page(bad_page); | |
964 | return page_to_pfn(bad_page); | |
965 | } | |
966 | ||
967 | return hva_to_pfn(kvm, addr); | |
968 | } | |
35149e21 AL |
969 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
970 | ||
506f0d6f MT |
971 | static unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn) |
972 | { | |
973 | return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE); | |
974 | } | |
975 | ||
976 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, | |
977 | struct kvm_memory_slot *slot, gfn_t gfn) | |
978 | { | |
979 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
980 | return hva_to_pfn(kvm, addr); | |
981 | } | |
982 | ||
35149e21 AL |
983 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
984 | { | |
2e2e3738 AL |
985 | pfn_t pfn; |
986 | ||
987 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 988 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
989 | return pfn_to_page(pfn); |
990 | ||
c77fb9dc | 991 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
992 | |
993 | get_page(bad_page); | |
994 | return bad_page; | |
954bbbc2 | 995 | } |
aab61cc0 | 996 | |
954bbbc2 AK |
997 | EXPORT_SYMBOL_GPL(gfn_to_page); |
998 | ||
b4231d61 IE |
999 | void kvm_release_page_clean(struct page *page) |
1000 | { | |
35149e21 | 1001 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1002 | } |
1003 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1004 | ||
35149e21 AL |
1005 | void kvm_release_pfn_clean(pfn_t pfn) |
1006 | { | |
c77fb9dc | 1007 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1008 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1009 | } |
1010 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1011 | ||
b4231d61 | 1012 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1013 | { |
35149e21 AL |
1014 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1015 | } | |
1016 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1017 | ||
1018 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1019 | { | |
1020 | kvm_set_pfn_dirty(pfn); | |
1021 | kvm_release_pfn_clean(pfn); | |
1022 | } | |
1023 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1024 | ||
1025 | void kvm_set_page_dirty(struct page *page) | |
1026 | { | |
1027 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1028 | } | |
1029 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1030 | ||
1031 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1032 | { | |
c77fb9dc | 1033 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1034 | struct page *page = pfn_to_page(pfn); |
1035 | if (!PageReserved(page)) | |
1036 | SetPageDirty(page); | |
1037 | } | |
8a7ae055 | 1038 | } |
35149e21 AL |
1039 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1040 | ||
1041 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1042 | { | |
c77fb9dc | 1043 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1044 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1045 | } |
1046 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1047 | ||
1048 | void kvm_get_pfn(pfn_t pfn) | |
1049 | { | |
c77fb9dc | 1050 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1051 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1052 | } |
1053 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1054 | |
195aefde IE |
1055 | static int next_segment(unsigned long len, int offset) |
1056 | { | |
1057 | if (len > PAGE_SIZE - offset) | |
1058 | return PAGE_SIZE - offset; | |
1059 | else | |
1060 | return len; | |
1061 | } | |
1062 | ||
1063 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1064 | int len) | |
1065 | { | |
e0506bcb IE |
1066 | int r; |
1067 | unsigned long addr; | |
195aefde | 1068 | |
e0506bcb IE |
1069 | addr = gfn_to_hva(kvm, gfn); |
1070 | if (kvm_is_error_hva(addr)) | |
1071 | return -EFAULT; | |
1072 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1073 | if (r) | |
195aefde | 1074 | return -EFAULT; |
195aefde IE |
1075 | return 0; |
1076 | } | |
1077 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1078 | ||
1079 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1080 | { | |
1081 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1082 | int seg; | |
1083 | int offset = offset_in_page(gpa); | |
1084 | int ret; | |
1085 | ||
1086 | while ((seg = next_segment(len, offset)) != 0) { | |
1087 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1088 | if (ret < 0) | |
1089 | return ret; | |
1090 | offset = 0; | |
1091 | len -= seg; | |
1092 | data += seg; | |
1093 | ++gfn; | |
1094 | } | |
1095 | return 0; | |
1096 | } | |
1097 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1098 | ||
7ec54588 MT |
1099 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1100 | unsigned long len) | |
1101 | { | |
1102 | int r; | |
1103 | unsigned long addr; | |
1104 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1105 | int offset = offset_in_page(gpa); | |
1106 | ||
1107 | addr = gfn_to_hva(kvm, gfn); | |
1108 | if (kvm_is_error_hva(addr)) | |
1109 | return -EFAULT; | |
0aac03f0 | 1110 | pagefault_disable(); |
7ec54588 | 1111 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1112 | pagefault_enable(); |
7ec54588 MT |
1113 | if (r) |
1114 | return -EFAULT; | |
1115 | return 0; | |
1116 | } | |
1117 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1118 | ||
195aefde IE |
1119 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1120 | int offset, int len) | |
1121 | { | |
e0506bcb IE |
1122 | int r; |
1123 | unsigned long addr; | |
195aefde | 1124 | |
e0506bcb IE |
1125 | addr = gfn_to_hva(kvm, gfn); |
1126 | if (kvm_is_error_hva(addr)) | |
1127 | return -EFAULT; | |
1128 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1129 | if (r) | |
195aefde | 1130 | return -EFAULT; |
195aefde IE |
1131 | mark_page_dirty(kvm, gfn); |
1132 | return 0; | |
1133 | } | |
1134 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1135 | ||
1136 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1137 | unsigned long len) | |
1138 | { | |
1139 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1140 | int seg; | |
1141 | int offset = offset_in_page(gpa); | |
1142 | int ret; | |
1143 | ||
1144 | while ((seg = next_segment(len, offset)) != 0) { | |
1145 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1146 | if (ret < 0) | |
1147 | return ret; | |
1148 | offset = 0; | |
1149 | len -= seg; | |
1150 | data += seg; | |
1151 | ++gfn; | |
1152 | } | |
1153 | return 0; | |
1154 | } | |
1155 | ||
1156 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1157 | { | |
3e021bf5 | 1158 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1159 | } |
1160 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1161 | ||
1162 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1163 | { | |
1164 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1165 | int seg; | |
1166 | int offset = offset_in_page(gpa); | |
1167 | int ret; | |
1168 | ||
1169 | while ((seg = next_segment(len, offset)) != 0) { | |
1170 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1171 | if (ret < 0) | |
1172 | return ret; | |
1173 | offset = 0; | |
1174 | len -= seg; | |
1175 | ++gfn; | |
1176 | } | |
1177 | return 0; | |
1178 | } | |
1179 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1180 | ||
6aa8b732 AK |
1181 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1182 | { | |
31389947 | 1183 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1184 | |
3b6fff19 | 1185 | gfn = unalias_gfn(kvm, gfn); |
2843099f | 1186 | memslot = gfn_to_memslot_unaliased(kvm, gfn); |
7e9d619d RR |
1187 | if (memslot && memslot->dirty_bitmap) { |
1188 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1189 | |
7e9d619d | 1190 | /* avoid RMW */ |
c8240bd6 AG |
1191 | if (!generic_test_le_bit(rel_gfn, memslot->dirty_bitmap)) |
1192 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1193 | } |
1194 | } | |
1195 | ||
b6958ce4 ED |
1196 | /* |
1197 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1198 | */ | |
8776e519 | 1199 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1200 | { |
e5c239cf MT |
1201 | DEFINE_WAIT(wait); |
1202 | ||
1203 | for (;;) { | |
1204 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1205 | ||
a1b37100 | 1206 | if (kvm_arch_vcpu_runnable(vcpu)) { |
d7690175 | 1207 | set_bit(KVM_REQ_UNHALT, &vcpu->requests); |
e5c239cf | 1208 | break; |
d7690175 | 1209 | } |
09cec754 GN |
1210 | if (kvm_cpu_has_pending_timer(vcpu)) |
1211 | break; | |
e5c239cf MT |
1212 | if (signal_pending(current)) |
1213 | break; | |
1214 | ||
b6958ce4 | 1215 | schedule(); |
b6958ce4 | 1216 | } |
d3bef15f | 1217 | |
e5c239cf | 1218 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1219 | } |
1220 | ||
6aa8b732 AK |
1221 | void kvm_resched(struct kvm_vcpu *vcpu) |
1222 | { | |
3fca0365 YD |
1223 | if (!need_resched()) |
1224 | return; | |
6aa8b732 | 1225 | cond_resched(); |
6aa8b732 AK |
1226 | } |
1227 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1228 | ||
d255f4f2 ZE |
1229 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu) |
1230 | { | |
1231 | ktime_t expires; | |
1232 | DEFINE_WAIT(wait); | |
1233 | ||
1234 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1235 | ||
1236 | /* Sleep for 100 us, and hope lock-holder got scheduled */ | |
1237 | expires = ktime_add_ns(ktime_get(), 100000UL); | |
1238 | schedule_hrtimeout(&expires, HRTIMER_MODE_ABS); | |
1239 | ||
1240 | finish_wait(&vcpu->wq, &wait); | |
1241 | } | |
1242 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1243 | ||
e4a533a4 | 1244 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1245 | { |
1246 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1247 | struct page *page; |
1248 | ||
e4a533a4 | 1249 | if (vmf->pgoff == 0) |
039576c0 | 1250 | page = virt_to_page(vcpu->run); |
09566765 | 1251 | #ifdef CONFIG_X86 |
e4a533a4 | 1252 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1253 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1254 | #endif |
1255 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1256 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1257 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1258 | #endif |
039576c0 | 1259 | else |
e4a533a4 | 1260 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1261 | get_page(page); |
e4a533a4 | 1262 | vmf->page = page; |
1263 | return 0; | |
9a2bb7f4 AK |
1264 | } |
1265 | ||
f0f37e2f | 1266 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1267 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1268 | }; |
1269 | ||
1270 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1271 | { | |
1272 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1273 | return 0; | |
1274 | } | |
1275 | ||
bccf2150 AK |
1276 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1277 | { | |
1278 | struct kvm_vcpu *vcpu = filp->private_data; | |
1279 | ||
66c0b394 | 1280 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1281 | return 0; |
1282 | } | |
1283 | ||
3d3aab1b | 1284 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1285 | .release = kvm_vcpu_release, |
1286 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1287 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1288 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1289 | }; |
1290 | ||
1291 | /* | |
1292 | * Allocates an inode for the vcpu. | |
1293 | */ | |
1294 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1295 | { | |
628ff7c1 | 1296 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1297 | } |
1298 | ||
c5ea7660 AK |
1299 | /* |
1300 | * Creates some virtual cpus. Good luck creating more than one. | |
1301 | */ | |
73880c80 | 1302 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1303 | { |
1304 | int r; | |
988a2cae | 1305 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1306 | |
73880c80 | 1307 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1308 | if (IS_ERR(vcpu)) |
1309 | return PTR_ERR(vcpu); | |
c5ea7660 | 1310 | |
15ad7146 AK |
1311 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1312 | ||
26e5215f AK |
1313 | r = kvm_arch_vcpu_setup(vcpu); |
1314 | if (r) | |
7d8fece6 | 1315 | return r; |
26e5215f | 1316 | |
11ec2804 | 1317 | mutex_lock(&kvm->lock); |
73880c80 GN |
1318 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1319 | r = -EINVAL; | |
e9b11c17 | 1320 | goto vcpu_destroy; |
fb3f0f51 | 1321 | } |
73880c80 | 1322 | |
988a2cae GN |
1323 | kvm_for_each_vcpu(r, v, kvm) |
1324 | if (v->vcpu_id == id) { | |
73880c80 GN |
1325 | r = -EEXIST; |
1326 | goto vcpu_destroy; | |
1327 | } | |
1328 | ||
1329 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1330 | |
fb3f0f51 | 1331 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1332 | kvm_get_kvm(kvm); |
bccf2150 | 1333 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1334 | if (r < 0) { |
1335 | kvm_put_kvm(kvm); | |
1336 | goto vcpu_destroy; | |
1337 | } | |
1338 | ||
1339 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1340 | smp_wmb(); | |
1341 | atomic_inc(&kvm->online_vcpus); | |
1342 | ||
1343 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1344 | if (kvm->bsp_vcpu_id == id) | |
1345 | kvm->bsp_vcpu = vcpu; | |
1346 | #endif | |
1347 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1348 | return r; |
39c3b86e | 1349 | |
e9b11c17 | 1350 | vcpu_destroy: |
7d8fece6 | 1351 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1352 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1353 | return r; |
1354 | } | |
1355 | ||
1961d276 AK |
1356 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1357 | { | |
1358 | if (sigset) { | |
1359 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1360 | vcpu->sigset_active = 1; | |
1361 | vcpu->sigset = *sigset; | |
1362 | } else | |
1363 | vcpu->sigset_active = 0; | |
1364 | return 0; | |
1365 | } | |
1366 | ||
bccf2150 AK |
1367 | static long kvm_vcpu_ioctl(struct file *filp, |
1368 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1369 | { |
bccf2150 | 1370 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1371 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1372 | int r; |
fa3795a7 DH |
1373 | struct kvm_fpu *fpu = NULL; |
1374 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1375 | |
6d4e4c4f AK |
1376 | if (vcpu->kvm->mm != current->mm) |
1377 | return -EIO; | |
6aa8b732 | 1378 | switch (ioctl) { |
9a2bb7f4 | 1379 | case KVM_RUN: |
f0fe5108 AK |
1380 | r = -EINVAL; |
1381 | if (arg) | |
1382 | goto out; | |
b6c7a5dc | 1383 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1384 | break; |
6aa8b732 | 1385 | case KVM_GET_REGS: { |
3e4bb3ac | 1386 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1387 | |
3e4bb3ac XZ |
1388 | r = -ENOMEM; |
1389 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1390 | if (!kvm_regs) | |
6aa8b732 | 1391 | goto out; |
3e4bb3ac XZ |
1392 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1393 | if (r) | |
1394 | goto out_free1; | |
6aa8b732 | 1395 | r = -EFAULT; |
3e4bb3ac XZ |
1396 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1397 | goto out_free1; | |
6aa8b732 | 1398 | r = 0; |
3e4bb3ac XZ |
1399 | out_free1: |
1400 | kfree(kvm_regs); | |
6aa8b732 AK |
1401 | break; |
1402 | } | |
1403 | case KVM_SET_REGS: { | |
3e4bb3ac | 1404 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1405 | |
3e4bb3ac XZ |
1406 | r = -ENOMEM; |
1407 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1408 | if (!kvm_regs) | |
6aa8b732 | 1409 | goto out; |
3e4bb3ac XZ |
1410 | r = -EFAULT; |
1411 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1412 | goto out_free2; | |
1413 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1414 | if (r) |
3e4bb3ac | 1415 | goto out_free2; |
6aa8b732 | 1416 | r = 0; |
3e4bb3ac XZ |
1417 | out_free2: |
1418 | kfree(kvm_regs); | |
6aa8b732 AK |
1419 | break; |
1420 | } | |
1421 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1422 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1423 | r = -ENOMEM; | |
1424 | if (!kvm_sregs) | |
1425 | goto out; | |
1426 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1427 | if (r) |
1428 | goto out; | |
1429 | r = -EFAULT; | |
fa3795a7 | 1430 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1431 | goto out; |
1432 | r = 0; | |
1433 | break; | |
1434 | } | |
1435 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1436 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1437 | r = -ENOMEM; | |
1438 | if (!kvm_sregs) | |
1439 | goto out; | |
6aa8b732 | 1440 | r = -EFAULT; |
fa3795a7 | 1441 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1442 | goto out; |
fa3795a7 | 1443 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1444 | if (r) |
1445 | goto out; | |
1446 | r = 0; | |
1447 | break; | |
1448 | } | |
62d9f0db MT |
1449 | case KVM_GET_MP_STATE: { |
1450 | struct kvm_mp_state mp_state; | |
1451 | ||
1452 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1453 | if (r) | |
1454 | goto out; | |
1455 | r = -EFAULT; | |
1456 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1457 | goto out; | |
1458 | r = 0; | |
1459 | break; | |
1460 | } | |
1461 | case KVM_SET_MP_STATE: { | |
1462 | struct kvm_mp_state mp_state; | |
1463 | ||
1464 | r = -EFAULT; | |
1465 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1466 | goto out; | |
1467 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1468 | if (r) | |
1469 | goto out; | |
1470 | r = 0; | |
1471 | break; | |
1472 | } | |
6aa8b732 AK |
1473 | case KVM_TRANSLATE: { |
1474 | struct kvm_translation tr; | |
1475 | ||
1476 | r = -EFAULT; | |
2f366987 | 1477 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1478 | goto out; |
8b006791 | 1479 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1480 | if (r) |
1481 | goto out; | |
1482 | r = -EFAULT; | |
2f366987 | 1483 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1484 | goto out; |
1485 | r = 0; | |
1486 | break; | |
1487 | } | |
d0bfb940 JK |
1488 | case KVM_SET_GUEST_DEBUG: { |
1489 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1490 | |
1491 | r = -EFAULT; | |
2f366987 | 1492 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1493 | goto out; |
d0bfb940 | 1494 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1495 | if (r) |
1496 | goto out; | |
1497 | r = 0; | |
1498 | break; | |
1499 | } | |
1961d276 AK |
1500 | case KVM_SET_SIGNAL_MASK: { |
1501 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1502 | struct kvm_signal_mask kvm_sigmask; | |
1503 | sigset_t sigset, *p; | |
1504 | ||
1505 | p = NULL; | |
1506 | if (argp) { | |
1507 | r = -EFAULT; | |
1508 | if (copy_from_user(&kvm_sigmask, argp, | |
1509 | sizeof kvm_sigmask)) | |
1510 | goto out; | |
1511 | r = -EINVAL; | |
1512 | if (kvm_sigmask.len != sizeof sigset) | |
1513 | goto out; | |
1514 | r = -EFAULT; | |
1515 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1516 | sizeof sigset)) | |
1517 | goto out; | |
1518 | p = &sigset; | |
1519 | } | |
1520 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1521 | break; | |
1522 | } | |
b8836737 | 1523 | case KVM_GET_FPU: { |
fa3795a7 DH |
1524 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1525 | r = -ENOMEM; | |
1526 | if (!fpu) | |
1527 | goto out; | |
1528 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1529 | if (r) |
1530 | goto out; | |
1531 | r = -EFAULT; | |
fa3795a7 | 1532 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1533 | goto out; |
1534 | r = 0; | |
1535 | break; | |
1536 | } | |
1537 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1538 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1539 | r = -ENOMEM; | |
1540 | if (!fpu) | |
1541 | goto out; | |
b8836737 | 1542 | r = -EFAULT; |
fa3795a7 | 1543 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1544 | goto out; |
fa3795a7 | 1545 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1546 | if (r) |
1547 | goto out; | |
1548 | r = 0; | |
1549 | break; | |
1550 | } | |
bccf2150 | 1551 | default: |
313a3dc7 | 1552 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1553 | } |
1554 | out: | |
fa3795a7 DH |
1555 | kfree(fpu); |
1556 | kfree(kvm_sregs); | |
bccf2150 AK |
1557 | return r; |
1558 | } | |
1559 | ||
1560 | static long kvm_vm_ioctl(struct file *filp, | |
1561 | unsigned int ioctl, unsigned long arg) | |
1562 | { | |
1563 | struct kvm *kvm = filp->private_data; | |
1564 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1565 | int r; |
bccf2150 | 1566 | |
6d4e4c4f AK |
1567 | if (kvm->mm != current->mm) |
1568 | return -EIO; | |
bccf2150 AK |
1569 | switch (ioctl) { |
1570 | case KVM_CREATE_VCPU: | |
1571 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1572 | if (r < 0) | |
1573 | goto out; | |
1574 | break; | |
6fc138d2 IE |
1575 | case KVM_SET_USER_MEMORY_REGION: { |
1576 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1577 | ||
1578 | r = -EFAULT; | |
1579 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1580 | sizeof kvm_userspace_mem)) | |
1581 | goto out; | |
1582 | ||
1583 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1584 | if (r) |
1585 | goto out; | |
1586 | break; | |
1587 | } | |
1588 | case KVM_GET_DIRTY_LOG: { | |
1589 | struct kvm_dirty_log log; | |
1590 | ||
1591 | r = -EFAULT; | |
2f366987 | 1592 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1593 | goto out; |
2c6f5df9 | 1594 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1595 | if (r) |
1596 | goto out; | |
1597 | break; | |
1598 | } | |
5f94c174 LV |
1599 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1600 | case KVM_REGISTER_COALESCED_MMIO: { | |
1601 | struct kvm_coalesced_mmio_zone zone; | |
1602 | r = -EFAULT; | |
1603 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1604 | goto out; | |
1605 | r = -ENXIO; | |
1606 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); | |
1607 | if (r) | |
1608 | goto out; | |
1609 | r = 0; | |
1610 | break; | |
1611 | } | |
1612 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1613 | struct kvm_coalesced_mmio_zone zone; | |
1614 | r = -EFAULT; | |
1615 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1616 | goto out; | |
1617 | r = -ENXIO; | |
1618 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); | |
1619 | if (r) | |
1620 | goto out; | |
1621 | r = 0; | |
1622 | break; | |
1623 | } | |
1624 | #endif | |
721eecbf GH |
1625 | case KVM_IRQFD: { |
1626 | struct kvm_irqfd data; | |
1627 | ||
1628 | r = -EFAULT; | |
1629 | if (copy_from_user(&data, argp, sizeof data)) | |
1630 | goto out; | |
1631 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1632 | break; | |
1633 | } | |
d34e6b17 GH |
1634 | case KVM_IOEVENTFD: { |
1635 | struct kvm_ioeventfd data; | |
1636 | ||
1637 | r = -EFAULT; | |
1638 | if (copy_from_user(&data, argp, sizeof data)) | |
1639 | goto out; | |
1640 | r = kvm_ioeventfd(kvm, &data); | |
1641 | break; | |
1642 | } | |
73880c80 GN |
1643 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1644 | case KVM_SET_BOOT_CPU_ID: | |
1645 | r = 0; | |
894a9c55 | 1646 | mutex_lock(&kvm->lock); |
73880c80 GN |
1647 | if (atomic_read(&kvm->online_vcpus) != 0) |
1648 | r = -EBUSY; | |
1649 | else | |
1650 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1651 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1652 | break; |
1653 | #endif | |
f17abe9a | 1654 | default: |
1fe779f8 | 1655 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1656 | if (r == -ENOTTY) |
1657 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1658 | } |
1659 | out: | |
1660 | return r; | |
1661 | } | |
1662 | ||
6ff5894c AB |
1663 | #ifdef CONFIG_COMPAT |
1664 | struct compat_kvm_dirty_log { | |
1665 | __u32 slot; | |
1666 | __u32 padding1; | |
1667 | union { | |
1668 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1669 | __u64 padding2; | |
1670 | }; | |
1671 | }; | |
1672 | ||
1673 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1674 | unsigned int ioctl, unsigned long arg) | |
1675 | { | |
1676 | struct kvm *kvm = filp->private_data; | |
1677 | int r; | |
1678 | ||
1679 | if (kvm->mm != current->mm) | |
1680 | return -EIO; | |
1681 | switch (ioctl) { | |
1682 | case KVM_GET_DIRTY_LOG: { | |
1683 | struct compat_kvm_dirty_log compat_log; | |
1684 | struct kvm_dirty_log log; | |
1685 | ||
1686 | r = -EFAULT; | |
1687 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1688 | sizeof(compat_log))) | |
1689 | goto out; | |
1690 | log.slot = compat_log.slot; | |
1691 | log.padding1 = compat_log.padding1; | |
1692 | log.padding2 = compat_log.padding2; | |
1693 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
1694 | ||
1695 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
1696 | if (r) | |
1697 | goto out; | |
1698 | break; | |
1699 | } | |
1700 | default: | |
1701 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
1702 | } | |
1703 | ||
1704 | out: | |
1705 | return r; | |
1706 | } | |
1707 | #endif | |
1708 | ||
e4a533a4 | 1709 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1710 | { |
777b3f49 MT |
1711 | struct page *page[1]; |
1712 | unsigned long addr; | |
1713 | int npages; | |
1714 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1715 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1716 | |
777b3f49 MT |
1717 | addr = gfn_to_hva(kvm, gfn); |
1718 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1719 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1720 | |
1721 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1722 | NULL); | |
1723 | if (unlikely(npages != 1)) | |
e4a533a4 | 1724 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1725 | |
1726 | vmf->page = page[0]; | |
e4a533a4 | 1727 | return 0; |
f17abe9a AK |
1728 | } |
1729 | ||
f0f37e2f | 1730 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 1731 | .fault = kvm_vm_fault, |
f17abe9a AK |
1732 | }; |
1733 | ||
1734 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1735 | { | |
1736 | vma->vm_ops = &kvm_vm_vm_ops; | |
1737 | return 0; | |
1738 | } | |
1739 | ||
3d3aab1b | 1740 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1741 | .release = kvm_vm_release, |
1742 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
1743 | #ifdef CONFIG_COMPAT |
1744 | .compat_ioctl = kvm_vm_compat_ioctl, | |
1745 | #endif | |
f17abe9a AK |
1746 | .mmap = kvm_vm_mmap, |
1747 | }; | |
1748 | ||
1749 | static int kvm_dev_ioctl_create_vm(void) | |
1750 | { | |
2030a42c | 1751 | int fd; |
f17abe9a AK |
1752 | struct kvm *kvm; |
1753 | ||
f17abe9a | 1754 | kvm = kvm_create_vm(); |
d6d28168 AK |
1755 | if (IS_ERR(kvm)) |
1756 | return PTR_ERR(kvm); | |
628ff7c1 | 1757 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2030a42c | 1758 | if (fd < 0) |
66c0b394 | 1759 | kvm_put_kvm(kvm); |
f17abe9a | 1760 | |
f17abe9a | 1761 | return fd; |
f17abe9a AK |
1762 | } |
1763 | ||
1a811b61 AK |
1764 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
1765 | { | |
1766 | switch (arg) { | |
ca9edaee | 1767 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 1768 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 1769 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
1770 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1771 | case KVM_CAP_SET_BOOT_CPU_ID: | |
1772 | #endif | |
a9c7399d | 1773 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 1774 | return 1; |
399ec807 AK |
1775 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
1776 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 1777 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 1778 | #endif |
1a811b61 AK |
1779 | default: |
1780 | break; | |
1781 | } | |
1782 | return kvm_dev_ioctl_check_extension(arg); | |
1783 | } | |
1784 | ||
f17abe9a AK |
1785 | static long kvm_dev_ioctl(struct file *filp, |
1786 | unsigned int ioctl, unsigned long arg) | |
1787 | { | |
07c45a36 | 1788 | long r = -EINVAL; |
f17abe9a AK |
1789 | |
1790 | switch (ioctl) { | |
1791 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1792 | r = -EINVAL; |
1793 | if (arg) | |
1794 | goto out; | |
f17abe9a AK |
1795 | r = KVM_API_VERSION; |
1796 | break; | |
1797 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1798 | r = -EINVAL; |
1799 | if (arg) | |
1800 | goto out; | |
f17abe9a AK |
1801 | r = kvm_dev_ioctl_create_vm(); |
1802 | break; | |
018d00d2 | 1803 | case KVM_CHECK_EXTENSION: |
1a811b61 | 1804 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 1805 | break; |
07c45a36 AK |
1806 | case KVM_GET_VCPU_MMAP_SIZE: |
1807 | r = -EINVAL; | |
1808 | if (arg) | |
1809 | goto out; | |
adb1ff46 AK |
1810 | r = PAGE_SIZE; /* struct kvm_run */ |
1811 | #ifdef CONFIG_X86 | |
1812 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
1813 | #endif |
1814 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1815 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 1816 | #endif |
07c45a36 | 1817 | break; |
d4c9ff2d FEL |
1818 | case KVM_TRACE_ENABLE: |
1819 | case KVM_TRACE_PAUSE: | |
1820 | case KVM_TRACE_DISABLE: | |
2023a29c | 1821 | r = -EOPNOTSUPP; |
d4c9ff2d | 1822 | break; |
6aa8b732 | 1823 | default: |
043405e1 | 1824 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1825 | } |
1826 | out: | |
1827 | return r; | |
1828 | } | |
1829 | ||
6aa8b732 | 1830 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1831 | .unlocked_ioctl = kvm_dev_ioctl, |
1832 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1833 | }; |
1834 | ||
1835 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1836 | KVM_MINOR, |
6aa8b732 AK |
1837 | "kvm", |
1838 | &kvm_chardev_ops, | |
1839 | }; | |
1840 | ||
1b6c0168 AK |
1841 | static void hardware_enable(void *junk) |
1842 | { | |
1843 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 1844 | int r; |
1b6c0168 | 1845 | |
7f59f492 | 1846 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1847 | return; |
10474ae8 | 1848 | |
7f59f492 | 1849 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
1850 | |
1851 | r = kvm_arch_hardware_enable(NULL); | |
1852 | ||
1853 | if (r) { | |
1854 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
1855 | atomic_inc(&hardware_enable_failed); | |
1856 | printk(KERN_INFO "kvm: enabling virtualization on " | |
1857 | "CPU%d failed\n", cpu); | |
1858 | } | |
1b6c0168 AK |
1859 | } |
1860 | ||
1861 | static void hardware_disable(void *junk) | |
1862 | { | |
1863 | int cpu = raw_smp_processor_id(); | |
1864 | ||
7f59f492 | 1865 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1866 | return; |
7f59f492 | 1867 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 1868 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1869 | } |
1870 | ||
10474ae8 AG |
1871 | static void hardware_disable_all_nolock(void) |
1872 | { | |
1873 | BUG_ON(!kvm_usage_count); | |
1874 | ||
1875 | kvm_usage_count--; | |
1876 | if (!kvm_usage_count) | |
1877 | on_each_cpu(hardware_disable, NULL, 1); | |
1878 | } | |
1879 | ||
1880 | static void hardware_disable_all(void) | |
1881 | { | |
1882 | spin_lock(&kvm_lock); | |
1883 | hardware_disable_all_nolock(); | |
1884 | spin_unlock(&kvm_lock); | |
1885 | } | |
1886 | ||
1887 | static int hardware_enable_all(void) | |
1888 | { | |
1889 | int r = 0; | |
1890 | ||
1891 | spin_lock(&kvm_lock); | |
1892 | ||
1893 | kvm_usage_count++; | |
1894 | if (kvm_usage_count == 1) { | |
1895 | atomic_set(&hardware_enable_failed, 0); | |
1896 | on_each_cpu(hardware_enable, NULL, 1); | |
1897 | ||
1898 | if (atomic_read(&hardware_enable_failed)) { | |
1899 | hardware_disable_all_nolock(); | |
1900 | r = -EBUSY; | |
1901 | } | |
1902 | } | |
1903 | ||
1904 | spin_unlock(&kvm_lock); | |
1905 | ||
1906 | return r; | |
1907 | } | |
1908 | ||
774c47f1 AK |
1909 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1910 | void *v) | |
1911 | { | |
1912 | int cpu = (long)v; | |
1913 | ||
10474ae8 AG |
1914 | if (!kvm_usage_count) |
1915 | return NOTIFY_OK; | |
1916 | ||
1a6f4d7f | 1917 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1918 | switch (val) { |
cec9ad27 | 1919 | case CPU_DYING: |
6ec8a856 AK |
1920 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1921 | cpu); | |
1922 | hardware_disable(NULL); | |
1923 | break; | |
774c47f1 | 1924 | case CPU_UP_CANCELED: |
43934a38 JK |
1925 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1926 | cpu); | |
8691e5a8 | 1927 | smp_call_function_single(cpu, hardware_disable, NULL, 1); |
774c47f1 | 1928 | break; |
43934a38 JK |
1929 | case CPU_ONLINE: |
1930 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1931 | cpu); | |
8691e5a8 | 1932 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
1933 | break; |
1934 | } | |
1935 | return NOTIFY_OK; | |
1936 | } | |
1937 | ||
4ecac3fd AK |
1938 | |
1939 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
1940 | { | |
1941 | if (kvm_rebooting) | |
1942 | /* spin while reset goes on */ | |
1943 | while (true) | |
1944 | ; | |
1945 | /* Fault while not rebooting. We want the trace. */ | |
1946 | BUG(); | |
1947 | } | |
1948 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
1949 | ||
9a2b85c6 | 1950 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1951 | void *v) |
9a2b85c6 | 1952 | { |
8e1c1815 SY |
1953 | /* |
1954 | * Some (well, at least mine) BIOSes hang on reboot if | |
1955 | * in vmx root mode. | |
1956 | * | |
1957 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
1958 | */ | |
1959 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1960 | kvm_rebooting = true; | |
1961 | on_each_cpu(hardware_disable, NULL, 1); | |
9a2b85c6 RR |
1962 | return NOTIFY_OK; |
1963 | } | |
1964 | ||
1965 | static struct notifier_block kvm_reboot_notifier = { | |
1966 | .notifier_call = kvm_reboot, | |
1967 | .priority = 0, | |
1968 | }; | |
1969 | ||
e93f8a0f | 1970 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
1971 | { |
1972 | int i; | |
1973 | ||
1974 | for (i = 0; i < bus->dev_count; i++) { | |
1975 | struct kvm_io_device *pos = bus->devs[i]; | |
1976 | ||
1977 | kvm_iodevice_destructor(pos); | |
1978 | } | |
e93f8a0f | 1979 | kfree(bus); |
2eeb2e94 GH |
1980 | } |
1981 | ||
bda9020e | 1982 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 1983 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 1984 | int len, const void *val) |
2eeb2e94 GH |
1985 | { |
1986 | int i; | |
e93f8a0f | 1987 | struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]); |
bda9020e MT |
1988 | for (i = 0; i < bus->dev_count; i++) |
1989 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
1990 | return 0; | |
1991 | return -EOPNOTSUPP; | |
1992 | } | |
2eeb2e94 | 1993 | |
bda9020e | 1994 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
1995 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
1996 | int len, void *val) | |
bda9020e MT |
1997 | { |
1998 | int i; | |
e93f8a0f MT |
1999 | struct kvm_io_bus *bus = rcu_dereference(kvm->buses[bus_idx]); |
2000 | ||
bda9020e MT |
2001 | for (i = 0; i < bus->dev_count; i++) |
2002 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
2003 | return 0; | |
2004 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
2005 | } |
2006 | ||
79fac95e | 2007 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2008 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2009 | struct kvm_io_device *dev) | |
6c474694 | 2010 | { |
e93f8a0f | 2011 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2012 | |
e93f8a0f | 2013 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2014 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2015 | return -ENOSPC; | |
2eeb2e94 | 2016 | |
e93f8a0f MT |
2017 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2018 | if (!new_bus) | |
2019 | return -ENOMEM; | |
2020 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2021 | new_bus->devs[new_bus->dev_count++] = dev; | |
2022 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2023 | synchronize_srcu_expedited(&kvm->srcu); | |
2024 | kfree(bus); | |
090b7aff GH |
2025 | |
2026 | return 0; | |
2027 | } | |
2028 | ||
79fac95e | 2029 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2030 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2031 | struct kvm_io_device *dev) | |
090b7aff | 2032 | { |
e93f8a0f MT |
2033 | int i, r; |
2034 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2035 | |
e93f8a0f MT |
2036 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2037 | if (!new_bus) | |
2038 | return -ENOMEM; | |
090b7aff | 2039 | |
e93f8a0f MT |
2040 | bus = kvm->buses[bus_idx]; |
2041 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2042 | ||
2043 | r = -ENOENT; | |
2044 | for (i = 0; i < new_bus->dev_count; i++) | |
2045 | if (new_bus->devs[i] == dev) { | |
2046 | r = 0; | |
2047 | new_bus->devs[i] = new_bus->devs[--new_bus->dev_count]; | |
090b7aff GH |
2048 | break; |
2049 | } | |
e93f8a0f MT |
2050 | |
2051 | if (r) { | |
2052 | kfree(new_bus); | |
2053 | return r; | |
2054 | } | |
2055 | ||
2056 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2057 | synchronize_srcu_expedited(&kvm->srcu); | |
2058 | kfree(bus); | |
2059 | return r; | |
2eeb2e94 GH |
2060 | } |
2061 | ||
774c47f1 AK |
2062 | static struct notifier_block kvm_cpu_notifier = { |
2063 | .notifier_call = kvm_cpu_hotplug, | |
2064 | .priority = 20, /* must be > scheduler priority */ | |
2065 | }; | |
2066 | ||
8b88b099 | 2067 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2068 | { |
2069 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2070 | struct kvm *kvm; |
2071 | ||
8b88b099 | 2072 | *val = 0; |
ba1389b7 AK |
2073 | spin_lock(&kvm_lock); |
2074 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2075 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2076 | spin_unlock(&kvm_lock); |
8b88b099 | 2077 | return 0; |
ba1389b7 AK |
2078 | } |
2079 | ||
2080 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2081 | ||
8b88b099 | 2082 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2083 | { |
2084 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2085 | struct kvm *kvm; |
2086 | struct kvm_vcpu *vcpu; | |
2087 | int i; | |
2088 | ||
8b88b099 | 2089 | *val = 0; |
1165f5fe AK |
2090 | spin_lock(&kvm_lock); |
2091 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
2092 | kvm_for_each_vcpu(i, vcpu, kvm) |
2093 | *val += *(u32 *)((void *)vcpu + offset); | |
2094 | ||
1165f5fe | 2095 | spin_unlock(&kvm_lock); |
8b88b099 | 2096 | return 0; |
1165f5fe AK |
2097 | } |
2098 | ||
ba1389b7 AK |
2099 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2100 | ||
828c0950 | 2101 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2102 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2103 | [KVM_STAT_VM] = &vm_stat_fops, | |
2104 | }; | |
1165f5fe | 2105 | |
a16b043c | 2106 | static void kvm_init_debug(void) |
6aa8b732 AK |
2107 | { |
2108 | struct kvm_stats_debugfs_item *p; | |
2109 | ||
76f7c879 | 2110 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2111 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2112 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2113 | (void *)(long)p->offset, |
ba1389b7 | 2114 | stat_fops[p->kind]); |
6aa8b732 AK |
2115 | } |
2116 | ||
2117 | static void kvm_exit_debug(void) | |
2118 | { | |
2119 | struct kvm_stats_debugfs_item *p; | |
2120 | ||
2121 | for (p = debugfs_entries; p->name; ++p) | |
2122 | debugfs_remove(p->dentry); | |
76f7c879 | 2123 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2124 | } |
2125 | ||
59ae6c6b AK |
2126 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2127 | { | |
10474ae8 AG |
2128 | if (kvm_usage_count) |
2129 | hardware_disable(NULL); | |
59ae6c6b AK |
2130 | return 0; |
2131 | } | |
2132 | ||
2133 | static int kvm_resume(struct sys_device *dev) | |
2134 | { | |
10474ae8 AG |
2135 | if (kvm_usage_count) |
2136 | hardware_enable(NULL); | |
59ae6c6b AK |
2137 | return 0; |
2138 | } | |
2139 | ||
2140 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2141 | .name = "kvm", |
59ae6c6b AK |
2142 | .suspend = kvm_suspend, |
2143 | .resume = kvm_resume, | |
2144 | }; | |
2145 | ||
2146 | static struct sys_device kvm_sysdev = { | |
2147 | .id = 0, | |
2148 | .cls = &kvm_sysdev_class, | |
2149 | }; | |
2150 | ||
cea7bb21 | 2151 | struct page *bad_page; |
35149e21 | 2152 | pfn_t bad_pfn; |
6aa8b732 | 2153 | |
15ad7146 AK |
2154 | static inline |
2155 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2156 | { | |
2157 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2158 | } | |
2159 | ||
2160 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2161 | { | |
2162 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2163 | ||
e9b11c17 | 2164 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2165 | } |
2166 | ||
2167 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2168 | struct task_struct *next) | |
2169 | { | |
2170 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2171 | ||
e9b11c17 | 2172 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2173 | } |
2174 | ||
f8c16bba | 2175 | int kvm_init(void *opaque, unsigned int vcpu_size, |
c16f862d | 2176 | struct module *module) |
6aa8b732 AK |
2177 | { |
2178 | int r; | |
002c7f7c | 2179 | int cpu; |
6aa8b732 | 2180 | |
f8c16bba ZX |
2181 | r = kvm_arch_init(opaque); |
2182 | if (r) | |
d2308784 | 2183 | goto out_fail; |
cb498ea2 ZX |
2184 | |
2185 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2186 | ||
2187 | if (bad_page == NULL) { | |
2188 | r = -ENOMEM; | |
2189 | goto out; | |
2190 | } | |
2191 | ||
35149e21 AL |
2192 | bad_pfn = page_to_pfn(bad_page); |
2193 | ||
8437a617 | 2194 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2195 | r = -ENOMEM; |
2196 | goto out_free_0; | |
2197 | } | |
2198 | ||
e9b11c17 | 2199 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2200 | if (r < 0) |
7f59f492 | 2201 | goto out_free_0a; |
6aa8b732 | 2202 | |
002c7f7c YS |
2203 | for_each_online_cpu(cpu) { |
2204 | smp_call_function_single(cpu, | |
e9b11c17 | 2205 | kvm_arch_check_processor_compat, |
8691e5a8 | 2206 | &r, 1); |
002c7f7c | 2207 | if (r < 0) |
d2308784 | 2208 | goto out_free_1; |
002c7f7c YS |
2209 | } |
2210 | ||
774c47f1 AK |
2211 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2212 | if (r) | |
d2308784 | 2213 | goto out_free_2; |
6aa8b732 AK |
2214 | register_reboot_notifier(&kvm_reboot_notifier); |
2215 | ||
59ae6c6b AK |
2216 | r = sysdev_class_register(&kvm_sysdev_class); |
2217 | if (r) | |
d2308784 | 2218 | goto out_free_3; |
59ae6c6b AK |
2219 | |
2220 | r = sysdev_register(&kvm_sysdev); | |
2221 | if (r) | |
d2308784 | 2222 | goto out_free_4; |
59ae6c6b | 2223 | |
c16f862d RR |
2224 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
2225 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
56919c5c JP |
2226 | __alignof__(struct kvm_vcpu), |
2227 | 0, NULL); | |
c16f862d RR |
2228 | if (!kvm_vcpu_cache) { |
2229 | r = -ENOMEM; | |
d2308784 | 2230 | goto out_free_5; |
c16f862d RR |
2231 | } |
2232 | ||
6aa8b732 | 2233 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2234 | kvm_vm_fops.owner = module; |
2235 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2236 | |
2237 | r = misc_register(&kvm_dev); | |
2238 | if (r) { | |
d77c26fc | 2239 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2240 | goto out_free; |
2241 | } | |
2242 | ||
15ad7146 AK |
2243 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2244 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2245 | ||
0ea4ed8e DW |
2246 | kvm_init_debug(); |
2247 | ||
c7addb90 | 2248 | return 0; |
6aa8b732 AK |
2249 | |
2250 | out_free: | |
c16f862d | 2251 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2252 | out_free_5: |
59ae6c6b | 2253 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2254 | out_free_4: |
59ae6c6b | 2255 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2256 | out_free_3: |
6aa8b732 | 2257 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2258 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2259 | out_free_2: |
d2308784 | 2260 | out_free_1: |
e9b11c17 | 2261 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2262 | out_free_0a: |
2263 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 ZX |
2264 | out_free_0: |
2265 | __free_page(bad_page); | |
ca45aaae | 2266 | out: |
f8c16bba | 2267 | kvm_arch_exit(); |
d2308784 | 2268 | out_fail: |
6aa8b732 AK |
2269 | return r; |
2270 | } | |
cb498ea2 | 2271 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2272 | |
cb498ea2 | 2273 | void kvm_exit(void) |
6aa8b732 | 2274 | { |
229456fc | 2275 | tracepoint_synchronize_unregister(); |
0ea4ed8e | 2276 | kvm_exit_debug(); |
6aa8b732 | 2277 | misc_deregister(&kvm_dev); |
c16f862d | 2278 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2279 | sysdev_unregister(&kvm_sysdev); |
2280 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2281 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2282 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2283 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2284 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2285 | kvm_arch_exit(); |
7f59f492 | 2286 | free_cpumask_var(cpus_hardware_enabled); |
cea7bb21 | 2287 | __free_page(bad_page); |
6aa8b732 | 2288 | } |
cb498ea2 | 2289 | EXPORT_SYMBOL_GPL(kvm_exit); |