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. | |
221d059d | 8 | * Copyright 2010 Red Hat, Inc. and/or its affilates. |
6aa8b732 AK |
9 | * |
10 | * Authors: | |
11 | * Avi Kivity <avi@qumranet.com> | |
12 | * Yaniv Kamay <yaniv@qumranet.com> | |
13 | * | |
14 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
15 | * the COPYING file in the top-level directory. | |
16 | * | |
17 | */ | |
18 | ||
e2174021 | 19 | #include "iodev.h" |
6aa8b732 | 20 | |
edf88417 | 21 | #include <linux/kvm_host.h> |
6aa8b732 AK |
22 | #include <linux/kvm.h> |
23 | #include <linux/module.h> | |
24 | #include <linux/errno.h> | |
6aa8b732 | 25 | #include <linux/percpu.h> |
6aa8b732 AK |
26 | #include <linux/mm.h> |
27 | #include <linux/miscdevice.h> | |
28 | #include <linux/vmalloc.h> | |
6aa8b732 | 29 | #include <linux/reboot.h> |
6aa8b732 AK |
30 | #include <linux/debugfs.h> |
31 | #include <linux/highmem.h> | |
32 | #include <linux/file.h> | |
59ae6c6b | 33 | #include <linux/sysdev.h> |
774c47f1 | 34 | #include <linux/cpu.h> |
e8edc6e0 | 35 | #include <linux/sched.h> |
d9e368d6 AK |
36 | #include <linux/cpumask.h> |
37 | #include <linux/smp.h> | |
d6d28168 | 38 | #include <linux/anon_inodes.h> |
04d2cc77 | 39 | #include <linux/profile.h> |
7aa81cc0 | 40 | #include <linux/kvm_para.h> |
6fc138d2 | 41 | #include <linux/pagemap.h> |
8d4e1288 | 42 | #include <linux/mman.h> |
35149e21 | 43 | #include <linux/swap.h> |
e56d532f | 44 | #include <linux/bitops.h> |
547de29e | 45 | #include <linux/spinlock.h> |
6ff5894c | 46 | #include <linux/compat.h> |
bc6678a3 | 47 | #include <linux/srcu.h> |
8f0b1ab6 | 48 | #include <linux/hugetlb.h> |
5a0e3ad6 | 49 | #include <linux/slab.h> |
6aa8b732 | 50 | |
e495606d | 51 | #include <asm/processor.h> |
e495606d AK |
52 | #include <asm/io.h> |
53 | #include <asm/uaccess.h> | |
3e021bf5 | 54 | #include <asm/pgtable.h> |
c8240bd6 | 55 | #include <asm-generic/bitops/le.h> |
6aa8b732 | 56 | |
5f94c174 | 57 | #include "coalesced_mmio.h" |
5f94c174 | 58 | |
229456fc MT |
59 | #define CREATE_TRACE_POINTS |
60 | #include <trace/events/kvm.h> | |
61 | ||
6aa8b732 AK |
62 | MODULE_AUTHOR("Qumranet"); |
63 | MODULE_LICENSE("GPL"); | |
64 | ||
fa40a821 MT |
65 | /* |
66 | * Ordering of locks: | |
67 | * | |
fae3a353 | 68 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
69 | */ |
70 | ||
e9b11c17 ZX |
71 | DEFINE_SPINLOCK(kvm_lock); |
72 | LIST_HEAD(vm_list); | |
133de902 | 73 | |
7f59f492 | 74 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
75 | static int kvm_usage_count = 0; |
76 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 77 | |
c16f862d RR |
78 | struct kmem_cache *kvm_vcpu_cache; |
79 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 80 | |
15ad7146 AK |
81 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
82 | ||
76f7c879 | 83 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 84 | |
bccf2150 AK |
85 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
86 | unsigned long arg); | |
10474ae8 AG |
87 | static int hardware_enable_all(void); |
88 | static void hardware_disable_all(void); | |
bccf2150 | 89 | |
e93f8a0f MT |
90 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
91 | ||
e8ba5d31 | 92 | static bool kvm_rebooting; |
4ecac3fd | 93 | |
54dee993 MT |
94 | static bool largepages_enabled = true; |
95 | ||
fa7bff8f GN |
96 | static struct page *hwpoison_page; |
97 | static pfn_t hwpoison_pfn; | |
bf998156 | 98 | |
c77fb9dc | 99 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 100 | { |
fc5659c8 JR |
101 | if (pfn_valid(pfn)) { |
102 | struct page *page = compound_head(pfn_to_page(pfn)); | |
103 | return PageReserved(page); | |
104 | } | |
cbff90a7 BAY |
105 | |
106 | return true; | |
107 | } | |
108 | ||
bccf2150 AK |
109 | /* |
110 | * Switches to specified vcpu, until a matching vcpu_put() | |
111 | */ | |
313a3dc7 | 112 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 113 | { |
15ad7146 AK |
114 | int cpu; |
115 | ||
bccf2150 | 116 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
117 | cpu = get_cpu(); |
118 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 119 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 120 | put_cpu(); |
6aa8b732 AK |
121 | } |
122 | ||
313a3dc7 | 123 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 124 | { |
15ad7146 | 125 | preempt_disable(); |
313a3dc7 | 126 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
127 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
128 | preempt_enable(); | |
6aa8b732 AK |
129 | mutex_unlock(&vcpu->mutex); |
130 | } | |
131 | ||
d9e368d6 AK |
132 | static void ack_flush(void *_completed) |
133 | { | |
d9e368d6 AK |
134 | } |
135 | ||
49846896 | 136 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 137 | { |
597a5f55 | 138 | int i, cpu, me; |
6ef7a1bc RR |
139 | cpumask_var_t cpus; |
140 | bool called = true; | |
d9e368d6 | 141 | struct kvm_vcpu *vcpu; |
d9e368d6 | 142 | |
79f55997 | 143 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 144 | |
70e335e1 | 145 | raw_spin_lock(&kvm->requests_lock); |
e601e3be | 146 | me = smp_processor_id(); |
988a2cae | 147 | kvm_for_each_vcpu(i, vcpu, kvm) { |
a8eeb04a | 148 | if (kvm_make_check_request(req, vcpu)) |
d9e368d6 AK |
149 | continue; |
150 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
151 | if (cpus != NULL && cpu != -1 && cpu != me) |
152 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 153 | } |
6ef7a1bc RR |
154 | if (unlikely(cpus == NULL)) |
155 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
156 | else if (!cpumask_empty(cpus)) | |
157 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
158 | else | |
159 | called = false; | |
70e335e1 | 160 | raw_spin_unlock(&kvm->requests_lock); |
6ef7a1bc | 161 | free_cpumask_var(cpus); |
49846896 | 162 | return called; |
d9e368d6 AK |
163 | } |
164 | ||
49846896 | 165 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 166 | { |
49846896 RR |
167 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
168 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
169 | } |
170 | ||
49846896 RR |
171 | void kvm_reload_remote_mmus(struct kvm *kvm) |
172 | { | |
173 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
174 | } | |
2e53d63a | 175 | |
fb3f0f51 RR |
176 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
177 | { | |
178 | struct page *page; | |
179 | int r; | |
180 | ||
181 | mutex_init(&vcpu->mutex); | |
182 | vcpu->cpu = -1; | |
fb3f0f51 RR |
183 | vcpu->kvm = kvm; |
184 | vcpu->vcpu_id = id; | |
b6958ce4 | 185 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
186 | |
187 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
188 | if (!page) { | |
189 | r = -ENOMEM; | |
190 | goto fail; | |
191 | } | |
192 | vcpu->run = page_address(page); | |
193 | ||
e9b11c17 | 194 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 195 | if (r < 0) |
e9b11c17 | 196 | goto fail_free_run; |
fb3f0f51 RR |
197 | return 0; |
198 | ||
fb3f0f51 RR |
199 | fail_free_run: |
200 | free_page((unsigned long)vcpu->run); | |
201 | fail: | |
76fafa5e | 202 | return r; |
fb3f0f51 RR |
203 | } |
204 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
205 | ||
206 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
207 | { | |
e9b11c17 | 208 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
209 | free_page((unsigned long)vcpu->run); |
210 | } | |
211 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
212 | ||
e930bffe AA |
213 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
214 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
215 | { | |
216 | return container_of(mn, struct kvm, mmu_notifier); | |
217 | } | |
218 | ||
219 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
220 | struct mm_struct *mm, | |
221 | unsigned long address) | |
222 | { | |
223 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 224 | int need_tlb_flush, idx; |
e930bffe AA |
225 | |
226 | /* | |
227 | * When ->invalidate_page runs, the linux pte has been zapped | |
228 | * already but the page is still allocated until | |
229 | * ->invalidate_page returns. So if we increase the sequence | |
230 | * here the kvm page fault will notice if the spte can't be | |
231 | * established because the page is going to be freed. If | |
232 | * instead the kvm page fault establishes the spte before | |
233 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
234 | * before returning. | |
235 | * | |
236 | * The sequence increase only need to be seen at spin_unlock | |
237 | * time, and not at spin_lock time. | |
238 | * | |
239 | * Increasing the sequence after the spin_unlock would be | |
240 | * unsafe because the kvm page fault could then establish the | |
241 | * pte after kvm_unmap_hva returned, without noticing the page | |
242 | * is going to be freed. | |
243 | */ | |
bc6678a3 | 244 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
245 | spin_lock(&kvm->mmu_lock); |
246 | kvm->mmu_notifier_seq++; | |
247 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
248 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 249 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
250 | |
251 | /* we've to flush the tlb before the pages can be freed */ | |
252 | if (need_tlb_flush) | |
253 | kvm_flush_remote_tlbs(kvm); | |
254 | ||
255 | } | |
256 | ||
3da0dd43 IE |
257 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
258 | struct mm_struct *mm, | |
259 | unsigned long address, | |
260 | pte_t pte) | |
261 | { | |
262 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 263 | int idx; |
3da0dd43 | 264 | |
bc6678a3 | 265 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
266 | spin_lock(&kvm->mmu_lock); |
267 | kvm->mmu_notifier_seq++; | |
268 | kvm_set_spte_hva(kvm, address, pte); | |
269 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 270 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
271 | } |
272 | ||
e930bffe AA |
273 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
274 | struct mm_struct *mm, | |
275 | unsigned long start, | |
276 | unsigned long end) | |
277 | { | |
278 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 279 | int need_tlb_flush = 0, idx; |
e930bffe | 280 | |
bc6678a3 | 281 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
282 | spin_lock(&kvm->mmu_lock); |
283 | /* | |
284 | * The count increase must become visible at unlock time as no | |
285 | * spte can be established without taking the mmu_lock and | |
286 | * count is also read inside the mmu_lock critical section. | |
287 | */ | |
288 | kvm->mmu_notifier_count++; | |
289 | for (; start < end; start += PAGE_SIZE) | |
290 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
291 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 292 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
293 | |
294 | /* we've to flush the tlb before the pages can be freed */ | |
295 | if (need_tlb_flush) | |
296 | kvm_flush_remote_tlbs(kvm); | |
297 | } | |
298 | ||
299 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
300 | struct mm_struct *mm, | |
301 | unsigned long start, | |
302 | unsigned long end) | |
303 | { | |
304 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
305 | ||
306 | spin_lock(&kvm->mmu_lock); | |
307 | /* | |
308 | * This sequence increase will notify the kvm page fault that | |
309 | * the page that is going to be mapped in the spte could have | |
310 | * been freed. | |
311 | */ | |
312 | kvm->mmu_notifier_seq++; | |
313 | /* | |
314 | * The above sequence increase must be visible before the | |
315 | * below count decrease but both values are read by the kvm | |
316 | * page fault under mmu_lock spinlock so we don't need to add | |
317 | * a smb_wmb() here in between the two. | |
318 | */ | |
319 | kvm->mmu_notifier_count--; | |
320 | spin_unlock(&kvm->mmu_lock); | |
321 | ||
322 | BUG_ON(kvm->mmu_notifier_count < 0); | |
323 | } | |
324 | ||
325 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
326 | struct mm_struct *mm, | |
327 | unsigned long address) | |
328 | { | |
329 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 330 | int young, idx; |
e930bffe | 331 | |
bc6678a3 | 332 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
333 | spin_lock(&kvm->mmu_lock); |
334 | young = kvm_age_hva(kvm, address); | |
335 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 336 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
337 | |
338 | if (young) | |
339 | kvm_flush_remote_tlbs(kvm); | |
340 | ||
341 | return young; | |
342 | } | |
343 | ||
85db06e5 MT |
344 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
345 | struct mm_struct *mm) | |
346 | { | |
347 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
348 | int idx; |
349 | ||
350 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 351 | kvm_arch_flush_shadow(kvm); |
eda2beda | 352 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
353 | } |
354 | ||
e930bffe AA |
355 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
356 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
357 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
358 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
359 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
3da0dd43 | 360 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 361 | .release = kvm_mmu_notifier_release, |
e930bffe | 362 | }; |
4c07b0a4 AK |
363 | |
364 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
365 | { | |
366 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
367 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
368 | } | |
369 | ||
370 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
371 | ||
372 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
373 | { | |
374 | return 0; | |
375 | } | |
376 | ||
e930bffe AA |
377 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
378 | ||
f17abe9a | 379 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 380 | { |
e93f8a0f | 381 | int r = 0, i; |
d19a9cd2 | 382 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 383 | |
d19a9cd2 ZX |
384 | if (IS_ERR(kvm)) |
385 | goto out; | |
10474ae8 AG |
386 | |
387 | r = hardware_enable_all(); | |
388 | if (r) | |
389 | goto out_err_nodisable; | |
390 | ||
75858a84 AK |
391 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
392 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 393 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 394 | #endif |
6aa8b732 | 395 | |
46a26bf5 MT |
396 | r = -ENOMEM; |
397 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
398 | if (!kvm->memslots) | |
399 | goto out_err; | |
bc6678a3 MT |
400 | if (init_srcu_struct(&kvm->srcu)) |
401 | goto out_err; | |
e93f8a0f MT |
402 | for (i = 0; i < KVM_NR_BUSES; i++) { |
403 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
404 | GFP_KERNEL); | |
405 | if (!kvm->buses[i]) { | |
406 | cleanup_srcu_struct(&kvm->srcu); | |
407 | goto out_err; | |
408 | } | |
409 | } | |
46a26bf5 | 410 | |
4c07b0a4 | 411 | r = kvm_init_mmu_notifier(kvm); |
283d0c65 | 412 | if (r) { |
bc6678a3 | 413 | cleanup_srcu_struct(&kvm->srcu); |
283d0c65 | 414 | goto out_err; |
e930bffe | 415 | } |
e930bffe | 416 | |
6d4e4c4f AK |
417 | kvm->mm = current->mm; |
418 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 419 | spin_lock_init(&kvm->mmu_lock); |
70e335e1 | 420 | raw_spin_lock_init(&kvm->requests_lock); |
d34e6b17 | 421 | kvm_eventfd_init(kvm); |
11ec2804 | 422 | mutex_init(&kvm->lock); |
60eead79 | 423 | mutex_init(&kvm->irq_lock); |
79fac95e | 424 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 425 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
426 | spin_lock(&kvm_lock); |
427 | list_add(&kvm->vm_list, &vm_list); | |
428 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 429 | out: |
f17abe9a | 430 | return kvm; |
10474ae8 AG |
431 | |
432 | out_err: | |
433 | hardware_disable_all(); | |
434 | out_err_nodisable: | |
e93f8a0f MT |
435 | for (i = 0; i < KVM_NR_BUSES; i++) |
436 | kfree(kvm->buses[i]); | |
46a26bf5 | 437 | kfree(kvm->memslots); |
10474ae8 AG |
438 | kfree(kvm); |
439 | return ERR_PTR(r); | |
f17abe9a AK |
440 | } |
441 | ||
6aa8b732 AK |
442 | /* |
443 | * Free any memory in @free but not in @dont. | |
444 | */ | |
445 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
446 | struct kvm_memory_slot *dont) | |
447 | { | |
ec04b260 JR |
448 | int i; |
449 | ||
290fc38d IE |
450 | if (!dont || free->rmap != dont->rmap) |
451 | vfree(free->rmap); | |
6aa8b732 AK |
452 | |
453 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
454 | vfree(free->dirty_bitmap); | |
455 | ||
ec04b260 JR |
456 | |
457 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
458 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
459 | vfree(free->lpage_info[i]); | |
460 | free->lpage_info[i] = NULL; | |
461 | } | |
462 | } | |
05da4558 | 463 | |
6aa8b732 | 464 | free->npages = 0; |
8b6d44c7 | 465 | free->dirty_bitmap = NULL; |
8d4e1288 | 466 | free->rmap = NULL; |
6aa8b732 AK |
467 | } |
468 | ||
d19a9cd2 | 469 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
470 | { |
471 | int i; | |
46a26bf5 MT |
472 | struct kvm_memslots *slots = kvm->memslots; |
473 | ||
474 | for (i = 0; i < slots->nmemslots; ++i) | |
475 | kvm_free_physmem_slot(&slots->memslots[i], NULL); | |
6aa8b732 | 476 | |
46a26bf5 | 477 | kfree(kvm->memslots); |
6aa8b732 AK |
478 | } |
479 | ||
f17abe9a AK |
480 | static void kvm_destroy_vm(struct kvm *kvm) |
481 | { | |
e93f8a0f | 482 | int i; |
6d4e4c4f AK |
483 | struct mm_struct *mm = kvm->mm; |
484 | ||
ad8ba2cd | 485 | kvm_arch_sync_events(kvm); |
133de902 AK |
486 | spin_lock(&kvm_lock); |
487 | list_del(&kvm->vm_list); | |
488 | spin_unlock(&kvm_lock); | |
399ec807 | 489 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
490 | for (i = 0; i < KVM_NR_BUSES; i++) |
491 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 492 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
493 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
494 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
495 | #else |
496 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 497 | #endif |
d19a9cd2 | 498 | kvm_arch_destroy_vm(kvm); |
10474ae8 | 499 | hardware_disable_all(); |
6d4e4c4f | 500 | mmdrop(mm); |
f17abe9a AK |
501 | } |
502 | ||
d39f13b0 IE |
503 | void kvm_get_kvm(struct kvm *kvm) |
504 | { | |
505 | atomic_inc(&kvm->users_count); | |
506 | } | |
507 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
508 | ||
509 | void kvm_put_kvm(struct kvm *kvm) | |
510 | { | |
511 | if (atomic_dec_and_test(&kvm->users_count)) | |
512 | kvm_destroy_vm(kvm); | |
513 | } | |
514 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
515 | ||
516 | ||
f17abe9a AK |
517 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
518 | { | |
519 | struct kvm *kvm = filp->private_data; | |
520 | ||
721eecbf GH |
521 | kvm_irqfd_release(kvm); |
522 | ||
d39f13b0 | 523 | kvm_put_kvm(kvm); |
6aa8b732 AK |
524 | return 0; |
525 | } | |
526 | ||
6aa8b732 AK |
527 | /* |
528 | * Allocate some memory and give it an address in the guest physical address | |
529 | * space. | |
530 | * | |
531 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 532 | * |
10589a46 | 533 | * Must be called holding mmap_sem for write. |
6aa8b732 | 534 | */ |
f78e0e2e SY |
535 | int __kvm_set_memory_region(struct kvm *kvm, |
536 | struct kvm_userspace_memory_region *mem, | |
537 | int user_alloc) | |
6aa8b732 | 538 | { |
bc6678a3 | 539 | int r, flush_shadow = 0; |
6aa8b732 | 540 | gfn_t base_gfn; |
28bcb112 HC |
541 | unsigned long npages; |
542 | unsigned long i; | |
6aa8b732 AK |
543 | struct kvm_memory_slot *memslot; |
544 | struct kvm_memory_slot old, new; | |
bc6678a3 | 545 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
546 | |
547 | r = -EINVAL; | |
548 | /* General sanity checks */ | |
549 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
550 | goto out; | |
551 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
552 | goto out; | |
e7cacd40 | 553 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 554 | goto out; |
e0d62c7f | 555 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
556 | goto out; |
557 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
558 | goto out; | |
559 | ||
46a26bf5 | 560 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
561 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
562 | npages = mem->memory_size >> PAGE_SHIFT; | |
563 | ||
660c22c4 TY |
564 | r = -EINVAL; |
565 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
566 | goto out; | |
567 | ||
6aa8b732 AK |
568 | if (!npages) |
569 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
570 | ||
6aa8b732 AK |
571 | new = old = *memslot; |
572 | ||
e36d96f7 | 573 | new.id = mem->slot; |
6aa8b732 AK |
574 | new.base_gfn = base_gfn; |
575 | new.npages = npages; | |
576 | new.flags = mem->flags; | |
577 | ||
578 | /* Disallow changing a memory slot's size. */ | |
579 | r = -EINVAL; | |
580 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 581 | goto out_free; |
6aa8b732 AK |
582 | |
583 | /* Check for overlaps */ | |
584 | r = -EEXIST; | |
585 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 586 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 587 | |
4cd481f6 | 588 | if (s == memslot || !s->npages) |
6aa8b732 AK |
589 | continue; |
590 | if (!((base_gfn + npages <= s->base_gfn) || | |
591 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 592 | goto out_free; |
6aa8b732 | 593 | } |
6aa8b732 | 594 | |
6aa8b732 AK |
595 | /* Free page dirty bitmap if unneeded */ |
596 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 597 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
598 | |
599 | r = -ENOMEM; | |
600 | ||
601 | /* Allocate if a slot is being created */ | |
eff0114a | 602 | #ifndef CONFIG_S390 |
8d4e1288 | 603 | if (npages && !new.rmap) { |
3bd89007 | 604 | new.rmap = vmalloc(npages * sizeof(*new.rmap)); |
290fc38d IE |
605 | |
606 | if (!new.rmap) | |
f78e0e2e | 607 | goto out_free; |
290fc38d | 608 | |
290fc38d | 609 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 610 | |
80b14b5b | 611 | new.user_alloc = user_alloc; |
bc6678a3 | 612 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 613 | } |
ec04b260 JR |
614 | if (!npages) |
615 | goto skip_lpage; | |
05da4558 | 616 | |
ec04b260 | 617 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
618 | unsigned long ugfn; |
619 | unsigned long j; | |
620 | int lpages; | |
ec04b260 | 621 | int level = i + 2; |
05da4558 | 622 | |
ec04b260 JR |
623 | /* Avoid unused variable warning if no large pages */ |
624 | (void)level; | |
625 | ||
626 | if (new.lpage_info[i]) | |
627 | continue; | |
628 | ||
82855413 JR |
629 | lpages = 1 + ((base_gfn + npages - 1) |
630 | >> KVM_HPAGE_GFN_SHIFT(level)); | |
631 | lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level); | |
ec04b260 JR |
632 | |
633 | new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i])); | |
634 | ||
635 | if (!new.lpage_info[i]) | |
05da4558 MT |
636 | goto out_free; |
637 | ||
ec04b260 JR |
638 | memset(new.lpage_info[i], 0, |
639 | lpages * sizeof(*new.lpage_info[i])); | |
05da4558 | 640 | |
82855413 | 641 | if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 642 | new.lpage_info[i][0].write_count = 1; |
82855413 | 643 | if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 644 | new.lpage_info[i][lpages - 1].write_count = 1; |
ac04527f AK |
645 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
646 | /* | |
647 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
648 | * other, or if explicitly asked to, disable large page |
649 | * support for this slot | |
ac04527f | 650 | */ |
ec04b260 | 651 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 652 | !largepages_enabled) |
ec04b260 JR |
653 | for (j = 0; j < lpages; ++j) |
654 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 655 | } |
6aa8b732 | 656 | |
ec04b260 JR |
657 | skip_lpage: |
658 | ||
6aa8b732 AK |
659 | /* Allocate page dirty bitmap if needed */ |
660 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
87bf6e7d | 661 | unsigned long dirty_bytes = kvm_dirty_bitmap_bytes(&new); |
6aa8b732 AK |
662 | |
663 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
664 | if (!new.dirty_bitmap) | |
f78e0e2e | 665 | goto out_free; |
6aa8b732 | 666 | memset(new.dirty_bitmap, 0, dirty_bytes); |
bc6678a3 | 667 | /* destroy any largepage mappings for dirty tracking */ |
e244584f | 668 | if (old.npages) |
bc6678a3 | 669 | flush_shadow = 1; |
6aa8b732 | 670 | } |
3eea8437 CB |
671 | #else /* not defined CONFIG_S390 */ |
672 | new.user_alloc = user_alloc; | |
673 | if (user_alloc) | |
674 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 675 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 676 | |
bc6678a3 MT |
677 | if (!npages) { |
678 | r = -ENOMEM; | |
679 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
680 | if (!slots) | |
681 | goto out_free; | |
682 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
683 | if (mem->slot >= slots->nmemslots) | |
684 | slots->nmemslots = mem->slot + 1; | |
685 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; | |
686 | ||
687 | old_memslots = kvm->memslots; | |
688 | rcu_assign_pointer(kvm->memslots, slots); | |
689 | synchronize_srcu_expedited(&kvm->srcu); | |
690 | /* From this point no new shadow pages pointing to a deleted | |
691 | * memslot will be created. | |
692 | * | |
693 | * validation of sp->gfn happens in: | |
694 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
695 | * - kvm_is_visible_gfn (mmu_check_roots) | |
696 | */ | |
34d4cb8f | 697 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
698 | kfree(old_memslots); |
699 | } | |
34d4cb8f | 700 | |
f7784b8e MT |
701 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
702 | if (r) | |
703 | goto out_free; | |
704 | ||
bc6678a3 MT |
705 | #ifdef CONFIG_DMAR |
706 | /* map the pages in iommu page table */ | |
707 | if (npages) { | |
708 | r = kvm_iommu_map_pages(kvm, &new); | |
709 | if (r) | |
710 | goto out_free; | |
711 | } | |
712 | #endif | |
604b38ac | 713 | |
bc6678a3 MT |
714 | r = -ENOMEM; |
715 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
716 | if (!slots) | |
717 | goto out_free; | |
718 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
719 | if (mem->slot >= slots->nmemslots) | |
720 | slots->nmemslots = mem->slot + 1; | |
721 | ||
722 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
723 | if (!npages) { | |
724 | new.rmap = NULL; | |
725 | new.dirty_bitmap = NULL; | |
726 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
727 | new.lpage_info[i] = NULL; | |
728 | } | |
729 | ||
730 | slots->memslots[mem->slot] = new; | |
731 | old_memslots = kvm->memslots; | |
732 | rcu_assign_pointer(kvm->memslots, slots); | |
733 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 734 | |
f7784b8e | 735 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 736 | |
bc6678a3 MT |
737 | kvm_free_physmem_slot(&old, &new); |
738 | kfree(old_memslots); | |
739 | ||
740 | if (flush_shadow) | |
741 | kvm_arch_flush_shadow(kvm); | |
742 | ||
6aa8b732 AK |
743 | return 0; |
744 | ||
f78e0e2e | 745 | out_free: |
6aa8b732 AK |
746 | kvm_free_physmem_slot(&new, &old); |
747 | out: | |
748 | return r; | |
210c7c4d IE |
749 | |
750 | } | |
f78e0e2e SY |
751 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
752 | ||
753 | int kvm_set_memory_region(struct kvm *kvm, | |
754 | struct kvm_userspace_memory_region *mem, | |
755 | int user_alloc) | |
756 | { | |
757 | int r; | |
758 | ||
79fac95e | 759 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 760 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 761 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
762 | return r; |
763 | } | |
210c7c4d IE |
764 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
765 | ||
1fe779f8 CO |
766 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
767 | struct | |
768 | kvm_userspace_memory_region *mem, | |
769 | int user_alloc) | |
210c7c4d | 770 | { |
e0d62c7f IE |
771 | if (mem->slot >= KVM_MEMORY_SLOTS) |
772 | return -EINVAL; | |
210c7c4d | 773 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
774 | } |
775 | ||
5bb064dc ZX |
776 | int kvm_get_dirty_log(struct kvm *kvm, |
777 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
778 | { |
779 | struct kvm_memory_slot *memslot; | |
780 | int r, i; | |
87bf6e7d | 781 | unsigned long n; |
6aa8b732 AK |
782 | unsigned long any = 0; |
783 | ||
6aa8b732 AK |
784 | r = -EINVAL; |
785 | if (log->slot >= KVM_MEMORY_SLOTS) | |
786 | goto out; | |
787 | ||
46a26bf5 | 788 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
789 | r = -ENOENT; |
790 | if (!memslot->dirty_bitmap) | |
791 | goto out; | |
792 | ||
87bf6e7d | 793 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 794 | |
cd1a4a98 | 795 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
796 | any = memslot->dirty_bitmap[i]; |
797 | ||
798 | r = -EFAULT; | |
799 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
800 | goto out; | |
801 | ||
5bb064dc ZX |
802 | if (any) |
803 | *is_dirty = 1; | |
6aa8b732 AK |
804 | |
805 | r = 0; | |
6aa8b732 | 806 | out: |
6aa8b732 AK |
807 | return r; |
808 | } | |
809 | ||
54dee993 MT |
810 | void kvm_disable_largepages(void) |
811 | { | |
812 | largepages_enabled = false; | |
813 | } | |
814 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
815 | ||
cea7bb21 IE |
816 | int is_error_page(struct page *page) |
817 | { | |
bf998156 | 818 | return page == bad_page || page == hwpoison_page; |
cea7bb21 IE |
819 | } |
820 | EXPORT_SYMBOL_GPL(is_error_page); | |
821 | ||
35149e21 AL |
822 | int is_error_pfn(pfn_t pfn) |
823 | { | |
bf998156 | 824 | return pfn == bad_pfn || pfn == hwpoison_pfn; |
35149e21 AL |
825 | } |
826 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
827 | ||
bf998156 HY |
828 | int is_hwpoison_pfn(pfn_t pfn) |
829 | { | |
830 | return pfn == hwpoison_pfn; | |
831 | } | |
832 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
833 | ||
f9d46eb0 IE |
834 | static inline unsigned long bad_hva(void) |
835 | { | |
836 | return PAGE_OFFSET; | |
837 | } | |
838 | ||
839 | int kvm_is_error_hva(unsigned long addr) | |
840 | { | |
841 | return addr == bad_hva(); | |
842 | } | |
843 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
844 | ||
a1f4d395 | 845 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
846 | { |
847 | int i; | |
90d83dc3 | 848 | struct kvm_memslots *slots = kvm_memslots(kvm); |
6aa8b732 | 849 | |
46a26bf5 MT |
850 | for (i = 0; i < slots->nmemslots; ++i) { |
851 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
852 | |
853 | if (gfn >= memslot->base_gfn | |
854 | && gfn < memslot->base_gfn + memslot->npages) | |
855 | return memslot; | |
856 | } | |
8b6d44c7 | 857 | return NULL; |
6aa8b732 | 858 | } |
a1f4d395 | 859 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 860 | |
e0d62c7f IE |
861 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
862 | { | |
863 | int i; | |
90d83dc3 | 864 | struct kvm_memslots *slots = kvm_memslots(kvm); |
e0d62c7f | 865 | |
e0d62c7f | 866 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 867 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 868 | |
bc6678a3 MT |
869 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
870 | continue; | |
871 | ||
e0d62c7f IE |
872 | if (gfn >= memslot->base_gfn |
873 | && gfn < memslot->base_gfn + memslot->npages) | |
874 | return 1; | |
875 | } | |
876 | return 0; | |
877 | } | |
878 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
879 | ||
8f0b1ab6 JR |
880 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
881 | { | |
882 | struct vm_area_struct *vma; | |
883 | unsigned long addr, size; | |
884 | ||
885 | size = PAGE_SIZE; | |
886 | ||
887 | addr = gfn_to_hva(kvm, gfn); | |
888 | if (kvm_is_error_hva(addr)) | |
889 | return PAGE_SIZE; | |
890 | ||
891 | down_read(¤t->mm->mmap_sem); | |
892 | vma = find_vma(current->mm, addr); | |
893 | if (!vma) | |
894 | goto out; | |
895 | ||
896 | size = vma_kernel_pagesize(vma); | |
897 | ||
898 | out: | |
899 | up_read(¤t->mm->mmap_sem); | |
900 | ||
901 | return size; | |
902 | } | |
903 | ||
bc6678a3 MT |
904 | int memslot_id(struct kvm *kvm, gfn_t gfn) |
905 | { | |
906 | int i; | |
90d83dc3 | 907 | struct kvm_memslots *slots = kvm_memslots(kvm); |
bc6678a3 MT |
908 | struct kvm_memory_slot *memslot = NULL; |
909 | ||
bc6678a3 MT |
910 | for (i = 0; i < slots->nmemslots; ++i) { |
911 | memslot = &slots->memslots[i]; | |
912 | ||
913 | if (gfn >= memslot->base_gfn | |
914 | && gfn < memslot->base_gfn + memslot->npages) | |
915 | break; | |
916 | } | |
917 | ||
918 | return memslot - slots->memslots; | |
919 | } | |
920 | ||
f5c98031 TY |
921 | static unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn) |
922 | { | |
923 | return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE; | |
924 | } | |
925 | ||
05da4558 | 926 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
927 | { |
928 | struct kvm_memory_slot *slot; | |
929 | ||
a1f4d395 | 930 | slot = gfn_to_memslot(kvm, gfn); |
bc6678a3 | 931 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 932 | return bad_hva(); |
f5c98031 | 933 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 934 | } |
0d150298 | 935 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 936 | |
506f0d6f | 937 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr) |
954bbbc2 | 938 | { |
8d4e1288 AL |
939 | struct page *page[1]; |
940 | int npages; | |
2e2e3738 | 941 | pfn_t pfn; |
954bbbc2 | 942 | |
60395224 AK |
943 | might_sleep(); |
944 | ||
4c2155ce | 945 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 946 | |
2e2e3738 AL |
947 | if (unlikely(npages != 1)) { |
948 | struct vm_area_struct *vma; | |
949 | ||
bbeb3406 | 950 | down_read(¤t->mm->mmap_sem); |
bf998156 | 951 | if (is_hwpoison_address(addr)) { |
bbeb3406 | 952 | up_read(¤t->mm->mmap_sem); |
bf998156 HY |
953 | get_page(hwpoison_page); |
954 | return page_to_pfn(hwpoison_page); | |
955 | } | |
956 | ||
2e2e3738 | 957 | vma = find_vma(current->mm, addr); |
4c2155ce | 958 | |
2e2e3738 AL |
959 | if (vma == NULL || addr < vma->vm_start || |
960 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 961 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
962 | get_page(bad_page); |
963 | return page_to_pfn(bad_page); | |
964 | } | |
965 | ||
966 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 967 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 968 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
969 | } else |
970 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 971 | |
2e2e3738 | 972 | return pfn; |
35149e21 AL |
973 | } |
974 | ||
506f0d6f MT |
975 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
976 | { | |
977 | unsigned long addr; | |
978 | ||
979 | addr = gfn_to_hva(kvm, gfn); | |
980 | if (kvm_is_error_hva(addr)) { | |
981 | get_page(bad_page); | |
982 | return page_to_pfn(bad_page); | |
983 | } | |
984 | ||
985 | return hva_to_pfn(kvm, addr); | |
986 | } | |
35149e21 AL |
987 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
988 | ||
506f0d6f MT |
989 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
990 | struct kvm_memory_slot *slot, gfn_t gfn) | |
991 | { | |
992 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
993 | return hva_to_pfn(kvm, addr); | |
994 | } | |
995 | ||
35149e21 AL |
996 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
997 | { | |
2e2e3738 AL |
998 | pfn_t pfn; |
999 | ||
1000 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1001 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1002 | return pfn_to_page(pfn); |
1003 | ||
c77fb9dc | 1004 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1005 | |
1006 | get_page(bad_page); | |
1007 | return bad_page; | |
954bbbc2 | 1008 | } |
aab61cc0 | 1009 | |
954bbbc2 AK |
1010 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1011 | ||
b4231d61 IE |
1012 | void kvm_release_page_clean(struct page *page) |
1013 | { | |
35149e21 | 1014 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1015 | } |
1016 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1017 | ||
35149e21 AL |
1018 | void kvm_release_pfn_clean(pfn_t pfn) |
1019 | { | |
c77fb9dc | 1020 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1021 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1022 | } |
1023 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1024 | ||
b4231d61 | 1025 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1026 | { |
35149e21 AL |
1027 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1028 | } | |
1029 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1030 | ||
1031 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1032 | { | |
1033 | kvm_set_pfn_dirty(pfn); | |
1034 | kvm_release_pfn_clean(pfn); | |
1035 | } | |
1036 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1037 | ||
1038 | void kvm_set_page_dirty(struct page *page) | |
1039 | { | |
1040 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1041 | } | |
1042 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1043 | ||
1044 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1045 | { | |
c77fb9dc | 1046 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1047 | struct page *page = pfn_to_page(pfn); |
1048 | if (!PageReserved(page)) | |
1049 | SetPageDirty(page); | |
1050 | } | |
8a7ae055 | 1051 | } |
35149e21 AL |
1052 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1053 | ||
1054 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1055 | { | |
c77fb9dc | 1056 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1057 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1058 | } |
1059 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1060 | ||
1061 | void kvm_get_pfn(pfn_t pfn) | |
1062 | { | |
c77fb9dc | 1063 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1064 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1065 | } |
1066 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1067 | |
195aefde IE |
1068 | static int next_segment(unsigned long len, int offset) |
1069 | { | |
1070 | if (len > PAGE_SIZE - offset) | |
1071 | return PAGE_SIZE - offset; | |
1072 | else | |
1073 | return len; | |
1074 | } | |
1075 | ||
1076 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1077 | int len) | |
1078 | { | |
e0506bcb IE |
1079 | int r; |
1080 | unsigned long addr; | |
195aefde | 1081 | |
e0506bcb IE |
1082 | addr = gfn_to_hva(kvm, gfn); |
1083 | if (kvm_is_error_hva(addr)) | |
1084 | return -EFAULT; | |
1085 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1086 | if (r) | |
195aefde | 1087 | return -EFAULT; |
195aefde IE |
1088 | return 0; |
1089 | } | |
1090 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1091 | ||
1092 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1093 | { | |
1094 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1095 | int seg; | |
1096 | int offset = offset_in_page(gpa); | |
1097 | int ret; | |
1098 | ||
1099 | while ((seg = next_segment(len, offset)) != 0) { | |
1100 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1101 | if (ret < 0) | |
1102 | return ret; | |
1103 | offset = 0; | |
1104 | len -= seg; | |
1105 | data += seg; | |
1106 | ++gfn; | |
1107 | } | |
1108 | return 0; | |
1109 | } | |
1110 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1111 | ||
7ec54588 MT |
1112 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1113 | unsigned long len) | |
1114 | { | |
1115 | int r; | |
1116 | unsigned long addr; | |
1117 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1118 | int offset = offset_in_page(gpa); | |
1119 | ||
1120 | addr = gfn_to_hva(kvm, gfn); | |
1121 | if (kvm_is_error_hva(addr)) | |
1122 | return -EFAULT; | |
0aac03f0 | 1123 | pagefault_disable(); |
7ec54588 | 1124 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1125 | pagefault_enable(); |
7ec54588 MT |
1126 | if (r) |
1127 | return -EFAULT; | |
1128 | return 0; | |
1129 | } | |
1130 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1131 | ||
195aefde IE |
1132 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1133 | int offset, int len) | |
1134 | { | |
e0506bcb IE |
1135 | int r; |
1136 | unsigned long addr; | |
195aefde | 1137 | |
e0506bcb IE |
1138 | addr = gfn_to_hva(kvm, gfn); |
1139 | if (kvm_is_error_hva(addr)) | |
1140 | return -EFAULT; | |
1141 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1142 | if (r) | |
195aefde | 1143 | return -EFAULT; |
195aefde IE |
1144 | mark_page_dirty(kvm, gfn); |
1145 | return 0; | |
1146 | } | |
1147 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1148 | ||
1149 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1150 | unsigned long len) | |
1151 | { | |
1152 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1153 | int seg; | |
1154 | int offset = offset_in_page(gpa); | |
1155 | int ret; | |
1156 | ||
1157 | while ((seg = next_segment(len, offset)) != 0) { | |
1158 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1159 | if (ret < 0) | |
1160 | return ret; | |
1161 | offset = 0; | |
1162 | len -= seg; | |
1163 | data += seg; | |
1164 | ++gfn; | |
1165 | } | |
1166 | return 0; | |
1167 | } | |
1168 | ||
1169 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1170 | { | |
3e021bf5 | 1171 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1172 | } |
1173 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1174 | ||
1175 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1176 | { | |
1177 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1178 | int seg; | |
1179 | int offset = offset_in_page(gpa); | |
1180 | int ret; | |
1181 | ||
1182 | while ((seg = next_segment(len, offset)) != 0) { | |
1183 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1184 | if (ret < 0) | |
1185 | return ret; | |
1186 | offset = 0; | |
1187 | len -= seg; | |
1188 | ++gfn; | |
1189 | } | |
1190 | return 0; | |
1191 | } | |
1192 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1193 | ||
6aa8b732 AK |
1194 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1195 | { | |
31389947 | 1196 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1197 | |
a1f4d395 | 1198 | memslot = gfn_to_memslot(kvm, gfn); |
7e9d619d RR |
1199 | if (memslot && memslot->dirty_bitmap) { |
1200 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1201 | |
d1476937 | 1202 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); |
6aa8b732 AK |
1203 | } |
1204 | } | |
1205 | ||
b6958ce4 ED |
1206 | /* |
1207 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1208 | */ | |
8776e519 | 1209 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1210 | { |
e5c239cf MT |
1211 | DEFINE_WAIT(wait); |
1212 | ||
1213 | for (;;) { | |
1214 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1215 | ||
a1b37100 | 1216 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1217 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1218 | break; |
d7690175 | 1219 | } |
09cec754 GN |
1220 | if (kvm_cpu_has_pending_timer(vcpu)) |
1221 | break; | |
e5c239cf MT |
1222 | if (signal_pending(current)) |
1223 | break; | |
1224 | ||
b6958ce4 | 1225 | schedule(); |
b6958ce4 | 1226 | } |
d3bef15f | 1227 | |
e5c239cf | 1228 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1229 | } |
1230 | ||
6aa8b732 AK |
1231 | void kvm_resched(struct kvm_vcpu *vcpu) |
1232 | { | |
3fca0365 YD |
1233 | if (!need_resched()) |
1234 | return; | |
6aa8b732 | 1235 | cond_resched(); |
6aa8b732 AK |
1236 | } |
1237 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1238 | ||
d255f4f2 ZE |
1239 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu) |
1240 | { | |
1241 | ktime_t expires; | |
1242 | DEFINE_WAIT(wait); | |
1243 | ||
1244 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1245 | ||
1246 | /* Sleep for 100 us, and hope lock-holder got scheduled */ | |
1247 | expires = ktime_add_ns(ktime_get(), 100000UL); | |
1248 | schedule_hrtimeout(&expires, HRTIMER_MODE_ABS); | |
1249 | ||
1250 | finish_wait(&vcpu->wq, &wait); | |
1251 | } | |
1252 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1253 | ||
e4a533a4 | 1254 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1255 | { |
1256 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1257 | struct page *page; |
1258 | ||
e4a533a4 | 1259 | if (vmf->pgoff == 0) |
039576c0 | 1260 | page = virt_to_page(vcpu->run); |
09566765 | 1261 | #ifdef CONFIG_X86 |
e4a533a4 | 1262 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1263 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1264 | #endif |
1265 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1266 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1267 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1268 | #endif |
039576c0 | 1269 | else |
e4a533a4 | 1270 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1271 | get_page(page); |
e4a533a4 | 1272 | vmf->page = page; |
1273 | return 0; | |
9a2bb7f4 AK |
1274 | } |
1275 | ||
f0f37e2f | 1276 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1277 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1278 | }; |
1279 | ||
1280 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1281 | { | |
1282 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1283 | return 0; | |
1284 | } | |
1285 | ||
bccf2150 AK |
1286 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1287 | { | |
1288 | struct kvm_vcpu *vcpu = filp->private_data; | |
1289 | ||
66c0b394 | 1290 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1291 | return 0; |
1292 | } | |
1293 | ||
3d3aab1b | 1294 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1295 | .release = kvm_vcpu_release, |
1296 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1297 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1298 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1299 | }; |
1300 | ||
1301 | /* | |
1302 | * Allocates an inode for the vcpu. | |
1303 | */ | |
1304 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1305 | { | |
628ff7c1 | 1306 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1307 | } |
1308 | ||
c5ea7660 AK |
1309 | /* |
1310 | * Creates some virtual cpus. Good luck creating more than one. | |
1311 | */ | |
73880c80 | 1312 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1313 | { |
1314 | int r; | |
988a2cae | 1315 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1316 | |
73880c80 | 1317 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1318 | if (IS_ERR(vcpu)) |
1319 | return PTR_ERR(vcpu); | |
c5ea7660 | 1320 | |
15ad7146 AK |
1321 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1322 | ||
26e5215f AK |
1323 | r = kvm_arch_vcpu_setup(vcpu); |
1324 | if (r) | |
7d8fece6 | 1325 | return r; |
26e5215f | 1326 | |
11ec2804 | 1327 | mutex_lock(&kvm->lock); |
73880c80 GN |
1328 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1329 | r = -EINVAL; | |
e9b11c17 | 1330 | goto vcpu_destroy; |
fb3f0f51 | 1331 | } |
73880c80 | 1332 | |
988a2cae GN |
1333 | kvm_for_each_vcpu(r, v, kvm) |
1334 | if (v->vcpu_id == id) { | |
73880c80 GN |
1335 | r = -EEXIST; |
1336 | goto vcpu_destroy; | |
1337 | } | |
1338 | ||
1339 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1340 | |
fb3f0f51 | 1341 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1342 | kvm_get_kvm(kvm); |
bccf2150 | 1343 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1344 | if (r < 0) { |
1345 | kvm_put_kvm(kvm); | |
1346 | goto vcpu_destroy; | |
1347 | } | |
1348 | ||
1349 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1350 | smp_wmb(); | |
1351 | atomic_inc(&kvm->online_vcpus); | |
1352 | ||
1353 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1354 | if (kvm->bsp_vcpu_id == id) | |
1355 | kvm->bsp_vcpu = vcpu; | |
1356 | #endif | |
1357 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1358 | return r; |
39c3b86e | 1359 | |
e9b11c17 | 1360 | vcpu_destroy: |
7d8fece6 | 1361 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1362 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1363 | return r; |
1364 | } | |
1365 | ||
1961d276 AK |
1366 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1367 | { | |
1368 | if (sigset) { | |
1369 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1370 | vcpu->sigset_active = 1; | |
1371 | vcpu->sigset = *sigset; | |
1372 | } else | |
1373 | vcpu->sigset_active = 0; | |
1374 | return 0; | |
1375 | } | |
1376 | ||
bccf2150 AK |
1377 | static long kvm_vcpu_ioctl(struct file *filp, |
1378 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1379 | { |
bccf2150 | 1380 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1381 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1382 | int r; |
fa3795a7 DH |
1383 | struct kvm_fpu *fpu = NULL; |
1384 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1385 | |
6d4e4c4f AK |
1386 | if (vcpu->kvm->mm != current->mm) |
1387 | return -EIO; | |
2122ff5e AK |
1388 | |
1389 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1390 | /* | |
1391 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1392 | * so vcpu_load() would break it. | |
1393 | */ | |
1394 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1395 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1396 | #endif | |
1397 | ||
1398 | ||
1399 | vcpu_load(vcpu); | |
6aa8b732 | 1400 | switch (ioctl) { |
9a2bb7f4 | 1401 | case KVM_RUN: |
f0fe5108 AK |
1402 | r = -EINVAL; |
1403 | if (arg) | |
1404 | goto out; | |
b6c7a5dc | 1405 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1406 | break; |
6aa8b732 | 1407 | case KVM_GET_REGS: { |
3e4bb3ac | 1408 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1409 | |
3e4bb3ac XZ |
1410 | r = -ENOMEM; |
1411 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1412 | if (!kvm_regs) | |
6aa8b732 | 1413 | goto out; |
3e4bb3ac XZ |
1414 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1415 | if (r) | |
1416 | goto out_free1; | |
6aa8b732 | 1417 | r = -EFAULT; |
3e4bb3ac XZ |
1418 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1419 | goto out_free1; | |
6aa8b732 | 1420 | r = 0; |
3e4bb3ac XZ |
1421 | out_free1: |
1422 | kfree(kvm_regs); | |
6aa8b732 AK |
1423 | break; |
1424 | } | |
1425 | case KVM_SET_REGS: { | |
3e4bb3ac | 1426 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1427 | |
3e4bb3ac XZ |
1428 | r = -ENOMEM; |
1429 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1430 | if (!kvm_regs) | |
6aa8b732 | 1431 | goto out; |
3e4bb3ac XZ |
1432 | r = -EFAULT; |
1433 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1434 | goto out_free2; | |
1435 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1436 | if (r) |
3e4bb3ac | 1437 | goto out_free2; |
6aa8b732 | 1438 | r = 0; |
3e4bb3ac XZ |
1439 | out_free2: |
1440 | kfree(kvm_regs); | |
6aa8b732 AK |
1441 | break; |
1442 | } | |
1443 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1444 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1445 | r = -ENOMEM; | |
1446 | if (!kvm_sregs) | |
1447 | goto out; | |
1448 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1449 | if (r) |
1450 | goto out; | |
1451 | r = -EFAULT; | |
fa3795a7 | 1452 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1453 | goto out; |
1454 | r = 0; | |
1455 | break; | |
1456 | } | |
1457 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1458 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1459 | r = -ENOMEM; | |
1460 | if (!kvm_sregs) | |
1461 | goto out; | |
6aa8b732 | 1462 | r = -EFAULT; |
fa3795a7 | 1463 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1464 | goto out; |
fa3795a7 | 1465 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1466 | if (r) |
1467 | goto out; | |
1468 | r = 0; | |
1469 | break; | |
1470 | } | |
62d9f0db MT |
1471 | case KVM_GET_MP_STATE: { |
1472 | struct kvm_mp_state mp_state; | |
1473 | ||
1474 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1475 | if (r) | |
1476 | goto out; | |
1477 | r = -EFAULT; | |
1478 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1479 | goto out; | |
1480 | r = 0; | |
1481 | break; | |
1482 | } | |
1483 | case KVM_SET_MP_STATE: { | |
1484 | struct kvm_mp_state mp_state; | |
1485 | ||
1486 | r = -EFAULT; | |
1487 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1488 | goto out; | |
1489 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1490 | if (r) | |
1491 | goto out; | |
1492 | r = 0; | |
1493 | break; | |
1494 | } | |
6aa8b732 AK |
1495 | case KVM_TRANSLATE: { |
1496 | struct kvm_translation tr; | |
1497 | ||
1498 | r = -EFAULT; | |
2f366987 | 1499 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1500 | goto out; |
8b006791 | 1501 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1502 | if (r) |
1503 | goto out; | |
1504 | r = -EFAULT; | |
2f366987 | 1505 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1506 | goto out; |
1507 | r = 0; | |
1508 | break; | |
1509 | } | |
d0bfb940 JK |
1510 | case KVM_SET_GUEST_DEBUG: { |
1511 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1512 | |
1513 | r = -EFAULT; | |
2f366987 | 1514 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1515 | goto out; |
d0bfb940 | 1516 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1517 | if (r) |
1518 | goto out; | |
1519 | r = 0; | |
1520 | break; | |
1521 | } | |
1961d276 AK |
1522 | case KVM_SET_SIGNAL_MASK: { |
1523 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1524 | struct kvm_signal_mask kvm_sigmask; | |
1525 | sigset_t sigset, *p; | |
1526 | ||
1527 | p = NULL; | |
1528 | if (argp) { | |
1529 | r = -EFAULT; | |
1530 | if (copy_from_user(&kvm_sigmask, argp, | |
1531 | sizeof kvm_sigmask)) | |
1532 | goto out; | |
1533 | r = -EINVAL; | |
1534 | if (kvm_sigmask.len != sizeof sigset) | |
1535 | goto out; | |
1536 | r = -EFAULT; | |
1537 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1538 | sizeof sigset)) | |
1539 | goto out; | |
1540 | p = &sigset; | |
1541 | } | |
376d41ff | 1542 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1543 | break; |
1544 | } | |
b8836737 | 1545 | case KVM_GET_FPU: { |
fa3795a7 DH |
1546 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1547 | r = -ENOMEM; | |
1548 | if (!fpu) | |
1549 | goto out; | |
1550 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1551 | if (r) |
1552 | goto out; | |
1553 | r = -EFAULT; | |
fa3795a7 | 1554 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1555 | goto out; |
1556 | r = 0; | |
1557 | break; | |
1558 | } | |
1559 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1560 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1561 | r = -ENOMEM; | |
1562 | if (!fpu) | |
1563 | goto out; | |
b8836737 | 1564 | r = -EFAULT; |
fa3795a7 | 1565 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1566 | goto out; |
fa3795a7 | 1567 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1568 | if (r) |
1569 | goto out; | |
1570 | r = 0; | |
1571 | break; | |
1572 | } | |
bccf2150 | 1573 | default: |
313a3dc7 | 1574 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1575 | } |
1576 | out: | |
2122ff5e | 1577 | vcpu_put(vcpu); |
fa3795a7 DH |
1578 | kfree(fpu); |
1579 | kfree(kvm_sregs); | |
bccf2150 AK |
1580 | return r; |
1581 | } | |
1582 | ||
1583 | static long kvm_vm_ioctl(struct file *filp, | |
1584 | unsigned int ioctl, unsigned long arg) | |
1585 | { | |
1586 | struct kvm *kvm = filp->private_data; | |
1587 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1588 | int r; |
bccf2150 | 1589 | |
6d4e4c4f AK |
1590 | if (kvm->mm != current->mm) |
1591 | return -EIO; | |
bccf2150 AK |
1592 | switch (ioctl) { |
1593 | case KVM_CREATE_VCPU: | |
1594 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1595 | if (r < 0) | |
1596 | goto out; | |
1597 | break; | |
6fc138d2 IE |
1598 | case KVM_SET_USER_MEMORY_REGION: { |
1599 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1600 | ||
1601 | r = -EFAULT; | |
1602 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1603 | sizeof kvm_userspace_mem)) | |
1604 | goto out; | |
1605 | ||
1606 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1607 | if (r) |
1608 | goto out; | |
1609 | break; | |
1610 | } | |
1611 | case KVM_GET_DIRTY_LOG: { | |
1612 | struct kvm_dirty_log log; | |
1613 | ||
1614 | r = -EFAULT; | |
2f366987 | 1615 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1616 | goto out; |
2c6f5df9 | 1617 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1618 | if (r) |
1619 | goto out; | |
1620 | break; | |
1621 | } | |
5f94c174 LV |
1622 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1623 | case KVM_REGISTER_COALESCED_MMIO: { | |
1624 | struct kvm_coalesced_mmio_zone zone; | |
1625 | r = -EFAULT; | |
1626 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1627 | goto out; | |
5f94c174 LV |
1628 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
1629 | if (r) | |
1630 | goto out; | |
1631 | r = 0; | |
1632 | break; | |
1633 | } | |
1634 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1635 | struct kvm_coalesced_mmio_zone zone; | |
1636 | r = -EFAULT; | |
1637 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1638 | goto out; | |
5f94c174 LV |
1639 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
1640 | if (r) | |
1641 | goto out; | |
1642 | r = 0; | |
1643 | break; | |
1644 | } | |
1645 | #endif | |
721eecbf GH |
1646 | case KVM_IRQFD: { |
1647 | struct kvm_irqfd data; | |
1648 | ||
1649 | r = -EFAULT; | |
1650 | if (copy_from_user(&data, argp, sizeof data)) | |
1651 | goto out; | |
1652 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1653 | break; | |
1654 | } | |
d34e6b17 GH |
1655 | case KVM_IOEVENTFD: { |
1656 | struct kvm_ioeventfd data; | |
1657 | ||
1658 | r = -EFAULT; | |
1659 | if (copy_from_user(&data, argp, sizeof data)) | |
1660 | goto out; | |
1661 | r = kvm_ioeventfd(kvm, &data); | |
1662 | break; | |
1663 | } | |
73880c80 GN |
1664 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1665 | case KVM_SET_BOOT_CPU_ID: | |
1666 | r = 0; | |
894a9c55 | 1667 | mutex_lock(&kvm->lock); |
73880c80 GN |
1668 | if (atomic_read(&kvm->online_vcpus) != 0) |
1669 | r = -EBUSY; | |
1670 | else | |
1671 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1672 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1673 | break; |
1674 | #endif | |
f17abe9a | 1675 | default: |
1fe779f8 | 1676 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1677 | if (r == -ENOTTY) |
1678 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1679 | } |
1680 | out: | |
1681 | return r; | |
1682 | } | |
1683 | ||
6ff5894c AB |
1684 | #ifdef CONFIG_COMPAT |
1685 | struct compat_kvm_dirty_log { | |
1686 | __u32 slot; | |
1687 | __u32 padding1; | |
1688 | union { | |
1689 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1690 | __u64 padding2; | |
1691 | }; | |
1692 | }; | |
1693 | ||
1694 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1695 | unsigned int ioctl, unsigned long arg) | |
1696 | { | |
1697 | struct kvm *kvm = filp->private_data; | |
1698 | int r; | |
1699 | ||
1700 | if (kvm->mm != current->mm) | |
1701 | return -EIO; | |
1702 | switch (ioctl) { | |
1703 | case KVM_GET_DIRTY_LOG: { | |
1704 | struct compat_kvm_dirty_log compat_log; | |
1705 | struct kvm_dirty_log log; | |
1706 | ||
1707 | r = -EFAULT; | |
1708 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1709 | sizeof(compat_log))) | |
1710 | goto out; | |
1711 | log.slot = compat_log.slot; | |
1712 | log.padding1 = compat_log.padding1; | |
1713 | log.padding2 = compat_log.padding2; | |
1714 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
1715 | ||
1716 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
1717 | if (r) | |
1718 | goto out; | |
1719 | break; | |
1720 | } | |
1721 | default: | |
1722 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
1723 | } | |
1724 | ||
1725 | out: | |
1726 | return r; | |
1727 | } | |
1728 | #endif | |
1729 | ||
e4a533a4 | 1730 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1731 | { |
777b3f49 MT |
1732 | struct page *page[1]; |
1733 | unsigned long addr; | |
1734 | int npages; | |
1735 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1736 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1737 | |
777b3f49 MT |
1738 | addr = gfn_to_hva(kvm, gfn); |
1739 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1740 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1741 | |
1742 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1743 | NULL); | |
1744 | if (unlikely(npages != 1)) | |
e4a533a4 | 1745 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1746 | |
1747 | vmf->page = page[0]; | |
e4a533a4 | 1748 | return 0; |
f17abe9a AK |
1749 | } |
1750 | ||
f0f37e2f | 1751 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 1752 | .fault = kvm_vm_fault, |
f17abe9a AK |
1753 | }; |
1754 | ||
1755 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1756 | { | |
1757 | vma->vm_ops = &kvm_vm_vm_ops; | |
1758 | return 0; | |
1759 | } | |
1760 | ||
3d3aab1b | 1761 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1762 | .release = kvm_vm_release, |
1763 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
1764 | #ifdef CONFIG_COMPAT |
1765 | .compat_ioctl = kvm_vm_compat_ioctl, | |
1766 | #endif | |
f17abe9a AK |
1767 | .mmap = kvm_vm_mmap, |
1768 | }; | |
1769 | ||
1770 | static int kvm_dev_ioctl_create_vm(void) | |
1771 | { | |
6ce5a090 | 1772 | int fd, r; |
f17abe9a AK |
1773 | struct kvm *kvm; |
1774 | ||
f17abe9a | 1775 | kvm = kvm_create_vm(); |
d6d28168 AK |
1776 | if (IS_ERR(kvm)) |
1777 | return PTR_ERR(kvm); | |
6ce5a090 TY |
1778 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1779 | r = kvm_coalesced_mmio_init(kvm); | |
1780 | if (r < 0) { | |
1781 | kvm_put_kvm(kvm); | |
1782 | return r; | |
1783 | } | |
1784 | #endif | |
628ff7c1 | 1785 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2030a42c | 1786 | if (fd < 0) |
66c0b394 | 1787 | kvm_put_kvm(kvm); |
f17abe9a | 1788 | |
f17abe9a | 1789 | return fd; |
f17abe9a AK |
1790 | } |
1791 | ||
1a811b61 AK |
1792 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
1793 | { | |
1794 | switch (arg) { | |
ca9edaee | 1795 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 1796 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 1797 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
1798 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1799 | case KVM_CAP_SET_BOOT_CPU_ID: | |
1800 | #endif | |
a9c7399d | 1801 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 1802 | return 1; |
399ec807 AK |
1803 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
1804 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 1805 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 1806 | #endif |
1a811b61 AK |
1807 | default: |
1808 | break; | |
1809 | } | |
1810 | return kvm_dev_ioctl_check_extension(arg); | |
1811 | } | |
1812 | ||
f17abe9a AK |
1813 | static long kvm_dev_ioctl(struct file *filp, |
1814 | unsigned int ioctl, unsigned long arg) | |
1815 | { | |
07c45a36 | 1816 | long r = -EINVAL; |
f17abe9a AK |
1817 | |
1818 | switch (ioctl) { | |
1819 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1820 | r = -EINVAL; |
1821 | if (arg) | |
1822 | goto out; | |
f17abe9a AK |
1823 | r = KVM_API_VERSION; |
1824 | break; | |
1825 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1826 | r = -EINVAL; |
1827 | if (arg) | |
1828 | goto out; | |
f17abe9a AK |
1829 | r = kvm_dev_ioctl_create_vm(); |
1830 | break; | |
018d00d2 | 1831 | case KVM_CHECK_EXTENSION: |
1a811b61 | 1832 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 1833 | break; |
07c45a36 AK |
1834 | case KVM_GET_VCPU_MMAP_SIZE: |
1835 | r = -EINVAL; | |
1836 | if (arg) | |
1837 | goto out; | |
adb1ff46 AK |
1838 | r = PAGE_SIZE; /* struct kvm_run */ |
1839 | #ifdef CONFIG_X86 | |
1840 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
1841 | #endif |
1842 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1843 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 1844 | #endif |
07c45a36 | 1845 | break; |
d4c9ff2d FEL |
1846 | case KVM_TRACE_ENABLE: |
1847 | case KVM_TRACE_PAUSE: | |
1848 | case KVM_TRACE_DISABLE: | |
2023a29c | 1849 | r = -EOPNOTSUPP; |
d4c9ff2d | 1850 | break; |
6aa8b732 | 1851 | default: |
043405e1 | 1852 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1853 | } |
1854 | out: | |
1855 | return r; | |
1856 | } | |
1857 | ||
6aa8b732 | 1858 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1859 | .unlocked_ioctl = kvm_dev_ioctl, |
1860 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1861 | }; |
1862 | ||
1863 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1864 | KVM_MINOR, |
6aa8b732 AK |
1865 | "kvm", |
1866 | &kvm_chardev_ops, | |
1867 | }; | |
1868 | ||
1b6c0168 AK |
1869 | static void hardware_enable(void *junk) |
1870 | { | |
1871 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 1872 | int r; |
1b6c0168 | 1873 | |
7f59f492 | 1874 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1875 | return; |
10474ae8 | 1876 | |
7f59f492 | 1877 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
1878 | |
1879 | r = kvm_arch_hardware_enable(NULL); | |
1880 | ||
1881 | if (r) { | |
1882 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
1883 | atomic_inc(&hardware_enable_failed); | |
1884 | printk(KERN_INFO "kvm: enabling virtualization on " | |
1885 | "CPU%d failed\n", cpu); | |
1886 | } | |
1b6c0168 AK |
1887 | } |
1888 | ||
1889 | static void hardware_disable(void *junk) | |
1890 | { | |
1891 | int cpu = raw_smp_processor_id(); | |
1892 | ||
7f59f492 | 1893 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1894 | return; |
7f59f492 | 1895 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 1896 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1897 | } |
1898 | ||
10474ae8 AG |
1899 | static void hardware_disable_all_nolock(void) |
1900 | { | |
1901 | BUG_ON(!kvm_usage_count); | |
1902 | ||
1903 | kvm_usage_count--; | |
1904 | if (!kvm_usage_count) | |
1905 | on_each_cpu(hardware_disable, NULL, 1); | |
1906 | } | |
1907 | ||
1908 | static void hardware_disable_all(void) | |
1909 | { | |
1910 | spin_lock(&kvm_lock); | |
1911 | hardware_disable_all_nolock(); | |
1912 | spin_unlock(&kvm_lock); | |
1913 | } | |
1914 | ||
1915 | static int hardware_enable_all(void) | |
1916 | { | |
1917 | int r = 0; | |
1918 | ||
1919 | spin_lock(&kvm_lock); | |
1920 | ||
1921 | kvm_usage_count++; | |
1922 | if (kvm_usage_count == 1) { | |
1923 | atomic_set(&hardware_enable_failed, 0); | |
1924 | on_each_cpu(hardware_enable, NULL, 1); | |
1925 | ||
1926 | if (atomic_read(&hardware_enable_failed)) { | |
1927 | hardware_disable_all_nolock(); | |
1928 | r = -EBUSY; | |
1929 | } | |
1930 | } | |
1931 | ||
1932 | spin_unlock(&kvm_lock); | |
1933 | ||
1934 | return r; | |
1935 | } | |
1936 | ||
774c47f1 AK |
1937 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1938 | void *v) | |
1939 | { | |
1940 | int cpu = (long)v; | |
1941 | ||
10474ae8 AG |
1942 | if (!kvm_usage_count) |
1943 | return NOTIFY_OK; | |
1944 | ||
1a6f4d7f | 1945 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1946 | switch (val) { |
cec9ad27 | 1947 | case CPU_DYING: |
6ec8a856 AK |
1948 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1949 | cpu); | |
1950 | hardware_disable(NULL); | |
1951 | break; | |
43934a38 JK |
1952 | case CPU_ONLINE: |
1953 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1954 | cpu); | |
8691e5a8 | 1955 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
1956 | break; |
1957 | } | |
1958 | return NOTIFY_OK; | |
1959 | } | |
1960 | ||
4ecac3fd AK |
1961 | |
1962 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
1963 | { | |
1964 | if (kvm_rebooting) | |
1965 | /* spin while reset goes on */ | |
1966 | while (true) | |
1967 | ; | |
1968 | /* Fault while not rebooting. We want the trace. */ | |
1969 | BUG(); | |
1970 | } | |
1971 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
1972 | ||
9a2b85c6 | 1973 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1974 | void *v) |
9a2b85c6 | 1975 | { |
8e1c1815 SY |
1976 | /* |
1977 | * Some (well, at least mine) BIOSes hang on reboot if | |
1978 | * in vmx root mode. | |
1979 | * | |
1980 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
1981 | */ | |
1982 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1983 | kvm_rebooting = true; | |
1984 | on_each_cpu(hardware_disable, NULL, 1); | |
9a2b85c6 RR |
1985 | return NOTIFY_OK; |
1986 | } | |
1987 | ||
1988 | static struct notifier_block kvm_reboot_notifier = { | |
1989 | .notifier_call = kvm_reboot, | |
1990 | .priority = 0, | |
1991 | }; | |
1992 | ||
e93f8a0f | 1993 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
1994 | { |
1995 | int i; | |
1996 | ||
1997 | for (i = 0; i < bus->dev_count; i++) { | |
1998 | struct kvm_io_device *pos = bus->devs[i]; | |
1999 | ||
2000 | kvm_iodevice_destructor(pos); | |
2001 | } | |
e93f8a0f | 2002 | kfree(bus); |
2eeb2e94 GH |
2003 | } |
2004 | ||
bda9020e | 2005 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2006 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2007 | int len, const void *val) |
2eeb2e94 GH |
2008 | { |
2009 | int i; | |
90d83dc3 LJ |
2010 | struct kvm_io_bus *bus; |
2011 | ||
2012 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
bda9020e MT |
2013 | for (i = 0; i < bus->dev_count; i++) |
2014 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
2015 | return 0; | |
2016 | return -EOPNOTSUPP; | |
2017 | } | |
2eeb2e94 | 2018 | |
bda9020e | 2019 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2020 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2021 | int len, void *val) | |
bda9020e MT |
2022 | { |
2023 | int i; | |
90d83dc3 | 2024 | struct kvm_io_bus *bus; |
e93f8a0f | 2025 | |
90d83dc3 | 2026 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
bda9020e MT |
2027 | for (i = 0; i < bus->dev_count; i++) |
2028 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
2029 | return 0; | |
2030 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
2031 | } |
2032 | ||
79fac95e | 2033 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2034 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2035 | struct kvm_io_device *dev) | |
6c474694 | 2036 | { |
e93f8a0f | 2037 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2038 | |
e93f8a0f | 2039 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2040 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2041 | return -ENOSPC; | |
2eeb2e94 | 2042 | |
e93f8a0f MT |
2043 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2044 | if (!new_bus) | |
2045 | return -ENOMEM; | |
2046 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2047 | new_bus->devs[new_bus->dev_count++] = dev; | |
2048 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2049 | synchronize_srcu_expedited(&kvm->srcu); | |
2050 | kfree(bus); | |
090b7aff GH |
2051 | |
2052 | return 0; | |
2053 | } | |
2054 | ||
79fac95e | 2055 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2056 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2057 | struct kvm_io_device *dev) | |
090b7aff | 2058 | { |
e93f8a0f MT |
2059 | int i, r; |
2060 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2061 | |
e93f8a0f MT |
2062 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2063 | if (!new_bus) | |
2064 | return -ENOMEM; | |
090b7aff | 2065 | |
e93f8a0f MT |
2066 | bus = kvm->buses[bus_idx]; |
2067 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2068 | ||
2069 | r = -ENOENT; | |
2070 | for (i = 0; i < new_bus->dev_count; i++) | |
2071 | if (new_bus->devs[i] == dev) { | |
2072 | r = 0; | |
2073 | new_bus->devs[i] = new_bus->devs[--new_bus->dev_count]; | |
090b7aff GH |
2074 | break; |
2075 | } | |
e93f8a0f MT |
2076 | |
2077 | if (r) { | |
2078 | kfree(new_bus); | |
2079 | return r; | |
2080 | } | |
2081 | ||
2082 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2083 | synchronize_srcu_expedited(&kvm->srcu); | |
2084 | kfree(bus); | |
2085 | return r; | |
2eeb2e94 GH |
2086 | } |
2087 | ||
774c47f1 AK |
2088 | static struct notifier_block kvm_cpu_notifier = { |
2089 | .notifier_call = kvm_cpu_hotplug, | |
2090 | .priority = 20, /* must be > scheduler priority */ | |
2091 | }; | |
2092 | ||
8b88b099 | 2093 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2094 | { |
2095 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2096 | struct kvm *kvm; |
2097 | ||
8b88b099 | 2098 | *val = 0; |
ba1389b7 AK |
2099 | spin_lock(&kvm_lock); |
2100 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2101 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2102 | spin_unlock(&kvm_lock); |
8b88b099 | 2103 | return 0; |
ba1389b7 AK |
2104 | } |
2105 | ||
2106 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2107 | ||
8b88b099 | 2108 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2109 | { |
2110 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2111 | struct kvm *kvm; |
2112 | struct kvm_vcpu *vcpu; | |
2113 | int i; | |
2114 | ||
8b88b099 | 2115 | *val = 0; |
1165f5fe AK |
2116 | spin_lock(&kvm_lock); |
2117 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
2118 | kvm_for_each_vcpu(i, vcpu, kvm) |
2119 | *val += *(u32 *)((void *)vcpu + offset); | |
2120 | ||
1165f5fe | 2121 | spin_unlock(&kvm_lock); |
8b88b099 | 2122 | return 0; |
1165f5fe AK |
2123 | } |
2124 | ||
ba1389b7 AK |
2125 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2126 | ||
828c0950 | 2127 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2128 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2129 | [KVM_STAT_VM] = &vm_stat_fops, | |
2130 | }; | |
1165f5fe | 2131 | |
a16b043c | 2132 | static void kvm_init_debug(void) |
6aa8b732 AK |
2133 | { |
2134 | struct kvm_stats_debugfs_item *p; | |
2135 | ||
76f7c879 | 2136 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2137 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2138 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2139 | (void *)(long)p->offset, |
ba1389b7 | 2140 | stat_fops[p->kind]); |
6aa8b732 AK |
2141 | } |
2142 | ||
2143 | static void kvm_exit_debug(void) | |
2144 | { | |
2145 | struct kvm_stats_debugfs_item *p; | |
2146 | ||
2147 | for (p = debugfs_entries; p->name; ++p) | |
2148 | debugfs_remove(p->dentry); | |
76f7c879 | 2149 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2150 | } |
2151 | ||
59ae6c6b AK |
2152 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2153 | { | |
10474ae8 AG |
2154 | if (kvm_usage_count) |
2155 | hardware_disable(NULL); | |
59ae6c6b AK |
2156 | return 0; |
2157 | } | |
2158 | ||
2159 | static int kvm_resume(struct sys_device *dev) | |
2160 | { | |
10474ae8 AG |
2161 | if (kvm_usage_count) |
2162 | hardware_enable(NULL); | |
59ae6c6b AK |
2163 | return 0; |
2164 | } | |
2165 | ||
2166 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2167 | .name = "kvm", |
59ae6c6b AK |
2168 | .suspend = kvm_suspend, |
2169 | .resume = kvm_resume, | |
2170 | }; | |
2171 | ||
2172 | static struct sys_device kvm_sysdev = { | |
2173 | .id = 0, | |
2174 | .cls = &kvm_sysdev_class, | |
2175 | }; | |
2176 | ||
cea7bb21 | 2177 | struct page *bad_page; |
35149e21 | 2178 | pfn_t bad_pfn; |
6aa8b732 | 2179 | |
15ad7146 AK |
2180 | static inline |
2181 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2182 | { | |
2183 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2184 | } | |
2185 | ||
2186 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2187 | { | |
2188 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2189 | ||
e9b11c17 | 2190 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2191 | } |
2192 | ||
2193 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2194 | struct task_struct *next) | |
2195 | { | |
2196 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2197 | ||
e9b11c17 | 2198 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2199 | } |
2200 | ||
0ee75bea | 2201 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2202 | struct module *module) |
6aa8b732 AK |
2203 | { |
2204 | int r; | |
002c7f7c | 2205 | int cpu; |
6aa8b732 | 2206 | |
f8c16bba ZX |
2207 | r = kvm_arch_init(opaque); |
2208 | if (r) | |
d2308784 | 2209 | goto out_fail; |
cb498ea2 ZX |
2210 | |
2211 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2212 | ||
2213 | if (bad_page == NULL) { | |
2214 | r = -ENOMEM; | |
2215 | goto out; | |
2216 | } | |
2217 | ||
35149e21 AL |
2218 | bad_pfn = page_to_pfn(bad_page); |
2219 | ||
bf998156 HY |
2220 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2221 | ||
2222 | if (hwpoison_page == NULL) { | |
2223 | r = -ENOMEM; | |
2224 | goto out_free_0; | |
2225 | } | |
2226 | ||
2227 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2228 | ||
8437a617 | 2229 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2230 | r = -ENOMEM; |
2231 | goto out_free_0; | |
2232 | } | |
2233 | ||
e9b11c17 | 2234 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2235 | if (r < 0) |
7f59f492 | 2236 | goto out_free_0a; |
6aa8b732 | 2237 | |
002c7f7c YS |
2238 | for_each_online_cpu(cpu) { |
2239 | smp_call_function_single(cpu, | |
e9b11c17 | 2240 | kvm_arch_check_processor_compat, |
8691e5a8 | 2241 | &r, 1); |
002c7f7c | 2242 | if (r < 0) |
d2308784 | 2243 | goto out_free_1; |
002c7f7c YS |
2244 | } |
2245 | ||
774c47f1 AK |
2246 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2247 | if (r) | |
d2308784 | 2248 | goto out_free_2; |
6aa8b732 AK |
2249 | register_reboot_notifier(&kvm_reboot_notifier); |
2250 | ||
59ae6c6b AK |
2251 | r = sysdev_class_register(&kvm_sysdev_class); |
2252 | if (r) | |
d2308784 | 2253 | goto out_free_3; |
59ae6c6b AK |
2254 | |
2255 | r = sysdev_register(&kvm_sysdev); | |
2256 | if (r) | |
d2308784 | 2257 | goto out_free_4; |
59ae6c6b | 2258 | |
c16f862d | 2259 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2260 | if (!vcpu_align) |
2261 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2262 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2263 | 0, NULL); |
c16f862d RR |
2264 | if (!kvm_vcpu_cache) { |
2265 | r = -ENOMEM; | |
d2308784 | 2266 | goto out_free_5; |
c16f862d RR |
2267 | } |
2268 | ||
6aa8b732 | 2269 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2270 | kvm_vm_fops.owner = module; |
2271 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2272 | |
2273 | r = misc_register(&kvm_dev); | |
2274 | if (r) { | |
d77c26fc | 2275 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2276 | goto out_free; |
2277 | } | |
2278 | ||
15ad7146 AK |
2279 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2280 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2281 | ||
0ea4ed8e DW |
2282 | kvm_init_debug(); |
2283 | ||
c7addb90 | 2284 | return 0; |
6aa8b732 AK |
2285 | |
2286 | out_free: | |
c16f862d | 2287 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2288 | out_free_5: |
59ae6c6b | 2289 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2290 | out_free_4: |
59ae6c6b | 2291 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2292 | out_free_3: |
6aa8b732 | 2293 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2294 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2295 | out_free_2: |
d2308784 | 2296 | out_free_1: |
e9b11c17 | 2297 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2298 | out_free_0a: |
2299 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2300 | out_free_0: |
bf998156 HY |
2301 | if (hwpoison_page) |
2302 | __free_page(hwpoison_page); | |
d2308784 | 2303 | __free_page(bad_page); |
ca45aaae | 2304 | out: |
f8c16bba | 2305 | kvm_arch_exit(); |
d2308784 | 2306 | out_fail: |
6aa8b732 AK |
2307 | return r; |
2308 | } | |
cb498ea2 | 2309 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2310 | |
cb498ea2 | 2311 | void kvm_exit(void) |
6aa8b732 | 2312 | { |
0ea4ed8e | 2313 | kvm_exit_debug(); |
6aa8b732 | 2314 | misc_deregister(&kvm_dev); |
c16f862d | 2315 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2316 | sysdev_unregister(&kvm_sysdev); |
2317 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2318 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2319 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2320 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2321 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2322 | kvm_arch_exit(); |
7f59f492 | 2323 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2324 | __free_page(hwpoison_page); |
cea7bb21 | 2325 | __free_page(bad_page); |
6aa8b732 | 2326 | } |
cb498ea2 | 2327 | EXPORT_SYMBOL_GPL(kvm_exit); |