]>
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. | |
9611c187 | 8 | * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
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" |
af585b92 | 58 | #include "async_pf.h" |
5f94c174 | 59 | |
229456fc MT |
60 | #define CREATE_TRACE_POINTS |
61 | #include <trace/events/kvm.h> | |
62 | ||
6aa8b732 AK |
63 | MODULE_AUTHOR("Qumranet"); |
64 | MODULE_LICENSE("GPL"); | |
65 | ||
fa40a821 MT |
66 | /* |
67 | * Ordering of locks: | |
68 | * | |
fae3a353 | 69 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
70 | */ |
71 | ||
e9b11c17 ZX |
72 | DEFINE_SPINLOCK(kvm_lock); |
73 | LIST_HEAD(vm_list); | |
133de902 | 74 | |
7f59f492 | 75 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
76 | static int kvm_usage_count = 0; |
77 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 78 | |
c16f862d RR |
79 | struct kmem_cache *kvm_vcpu_cache; |
80 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 81 | |
15ad7146 AK |
82 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
83 | ||
76f7c879 | 84 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 85 | |
bccf2150 AK |
86 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
87 | unsigned long arg); | |
10474ae8 AG |
88 | static int hardware_enable_all(void); |
89 | static void hardware_disable_all(void); | |
bccf2150 | 90 | |
e93f8a0f MT |
91 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
92 | ||
b7c4145b AK |
93 | bool kvm_rebooting; |
94 | EXPORT_SYMBOL_GPL(kvm_rebooting); | |
4ecac3fd | 95 | |
54dee993 MT |
96 | static bool largepages_enabled = true; |
97 | ||
fa7bff8f GN |
98 | static struct page *hwpoison_page; |
99 | static pfn_t hwpoison_pfn; | |
bf998156 | 100 | |
edba23e5 GN |
101 | static struct page *fault_page; |
102 | static pfn_t fault_pfn; | |
103 | ||
c77fb9dc | 104 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 105 | { |
fc5659c8 | 106 | if (pfn_valid(pfn)) { |
936a5fe6 AA |
107 | struct page *head; |
108 | struct page *tail = pfn_to_page(pfn); | |
109 | head = compound_head(tail); | |
110 | if (head != tail) { | |
111 | smp_rmb(); | |
112 | /* | |
113 | * head may be a dangling pointer. | |
114 | * __split_huge_page_refcount clears PageTail | |
115 | * before overwriting first_page, so if | |
116 | * PageTail is still there it means the head | |
117 | * pointer isn't dangling. | |
118 | */ | |
119 | if (PageTail(tail)) { | |
120 | /* | |
121 | * the "head" is not a dangling | |
122 | * pointer but the hugepage may have | |
123 | * been splitted from under us (and we | |
124 | * may not hold a reference count on | |
125 | * the head page so it can be reused | |
126 | * before we run PageReferenced), so | |
127 | * we've to recheck PageTail before | |
128 | * returning what we just read. | |
129 | */ | |
130 | int reserved = PageReserved(head); | |
131 | smp_rmb(); | |
132 | if (PageTail(tail)) | |
133 | return reserved; | |
134 | } | |
135 | } | |
136 | return PageReserved(tail); | |
fc5659c8 | 137 | } |
cbff90a7 BAY |
138 | |
139 | return true; | |
140 | } | |
141 | ||
bccf2150 AK |
142 | /* |
143 | * Switches to specified vcpu, until a matching vcpu_put() | |
144 | */ | |
313a3dc7 | 145 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 146 | { |
15ad7146 AK |
147 | int cpu; |
148 | ||
bccf2150 | 149 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
150 | cpu = get_cpu(); |
151 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 152 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 153 | put_cpu(); |
6aa8b732 AK |
154 | } |
155 | ||
313a3dc7 | 156 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 157 | { |
15ad7146 | 158 | preempt_disable(); |
313a3dc7 | 159 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
160 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
161 | preempt_enable(); | |
6aa8b732 AK |
162 | mutex_unlock(&vcpu->mutex); |
163 | } | |
164 | ||
d9e368d6 AK |
165 | static void ack_flush(void *_completed) |
166 | { | |
d9e368d6 AK |
167 | } |
168 | ||
49846896 | 169 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 170 | { |
597a5f55 | 171 | int i, cpu, me; |
6ef7a1bc RR |
172 | cpumask_var_t cpus; |
173 | bool called = true; | |
d9e368d6 | 174 | struct kvm_vcpu *vcpu; |
d9e368d6 | 175 | |
79f55997 | 176 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 177 | |
70e335e1 | 178 | raw_spin_lock(&kvm->requests_lock); |
e601e3be | 179 | me = smp_processor_id(); |
988a2cae | 180 | kvm_for_each_vcpu(i, vcpu, kvm) { |
a8eeb04a | 181 | if (kvm_make_check_request(req, vcpu)) |
d9e368d6 AK |
182 | continue; |
183 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
184 | if (cpus != NULL && cpu != -1 && cpu != me) |
185 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 186 | } |
6ef7a1bc RR |
187 | if (unlikely(cpus == NULL)) |
188 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
189 | else if (!cpumask_empty(cpus)) | |
190 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
191 | else | |
192 | called = false; | |
70e335e1 | 193 | raw_spin_unlock(&kvm->requests_lock); |
6ef7a1bc | 194 | free_cpumask_var(cpus); |
49846896 | 195 | return called; |
d9e368d6 AK |
196 | } |
197 | ||
49846896 | 198 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 199 | { |
a4ee1ca4 XG |
200 | int dirty_count = kvm->tlbs_dirty; |
201 | ||
202 | smp_mb(); | |
49846896 RR |
203 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
204 | ++kvm->stat.remote_tlb_flush; | |
a4ee1ca4 | 205 | cmpxchg(&kvm->tlbs_dirty, dirty_count, 0); |
2e53d63a MT |
206 | } |
207 | ||
49846896 RR |
208 | void kvm_reload_remote_mmus(struct kvm *kvm) |
209 | { | |
210 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
211 | } | |
2e53d63a | 212 | |
fb3f0f51 RR |
213 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
214 | { | |
215 | struct page *page; | |
216 | int r; | |
217 | ||
218 | mutex_init(&vcpu->mutex); | |
219 | vcpu->cpu = -1; | |
fb3f0f51 RR |
220 | vcpu->kvm = kvm; |
221 | vcpu->vcpu_id = id; | |
b6958ce4 | 222 | init_waitqueue_head(&vcpu->wq); |
af585b92 | 223 | kvm_async_pf_vcpu_init(vcpu); |
fb3f0f51 RR |
224 | |
225 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
226 | if (!page) { | |
227 | r = -ENOMEM; | |
228 | goto fail; | |
229 | } | |
230 | vcpu->run = page_address(page); | |
231 | ||
e9b11c17 | 232 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 233 | if (r < 0) |
e9b11c17 | 234 | goto fail_free_run; |
fb3f0f51 RR |
235 | return 0; |
236 | ||
fb3f0f51 RR |
237 | fail_free_run: |
238 | free_page((unsigned long)vcpu->run); | |
239 | fail: | |
76fafa5e | 240 | return r; |
fb3f0f51 RR |
241 | } |
242 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
243 | ||
244 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
245 | { | |
e9b11c17 | 246 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
247 | free_page((unsigned long)vcpu->run); |
248 | } | |
249 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
250 | ||
e930bffe AA |
251 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
252 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
253 | { | |
254 | return container_of(mn, struct kvm, mmu_notifier); | |
255 | } | |
256 | ||
257 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
258 | struct mm_struct *mm, | |
259 | unsigned long address) | |
260 | { | |
261 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 262 | int need_tlb_flush, idx; |
e930bffe AA |
263 | |
264 | /* | |
265 | * When ->invalidate_page runs, the linux pte has been zapped | |
266 | * already but the page is still allocated until | |
267 | * ->invalidate_page returns. So if we increase the sequence | |
268 | * here the kvm page fault will notice if the spte can't be | |
269 | * established because the page is going to be freed. If | |
270 | * instead the kvm page fault establishes the spte before | |
271 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
272 | * before returning. | |
273 | * | |
274 | * The sequence increase only need to be seen at spin_unlock | |
275 | * time, and not at spin_lock time. | |
276 | * | |
277 | * Increasing the sequence after the spin_unlock would be | |
278 | * unsafe because the kvm page fault could then establish the | |
279 | * pte after kvm_unmap_hva returned, without noticing the page | |
280 | * is going to be freed. | |
281 | */ | |
bc6678a3 | 282 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
283 | spin_lock(&kvm->mmu_lock); |
284 | kvm->mmu_notifier_seq++; | |
a4ee1ca4 | 285 | need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty; |
e930bffe | 286 | spin_unlock(&kvm->mmu_lock); |
bc6678a3 | 287 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
288 | |
289 | /* we've to flush the tlb before the pages can be freed */ | |
290 | if (need_tlb_flush) | |
291 | kvm_flush_remote_tlbs(kvm); | |
292 | ||
293 | } | |
294 | ||
3da0dd43 IE |
295 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
296 | struct mm_struct *mm, | |
297 | unsigned long address, | |
298 | pte_t pte) | |
299 | { | |
300 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 301 | int idx; |
3da0dd43 | 302 | |
bc6678a3 | 303 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
304 | spin_lock(&kvm->mmu_lock); |
305 | kvm->mmu_notifier_seq++; | |
306 | kvm_set_spte_hva(kvm, address, pte); | |
307 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 308 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
309 | } |
310 | ||
e930bffe AA |
311 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
312 | struct mm_struct *mm, | |
313 | unsigned long start, | |
314 | unsigned long end) | |
315 | { | |
316 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 317 | int need_tlb_flush = 0, idx; |
e930bffe | 318 | |
bc6678a3 | 319 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
320 | spin_lock(&kvm->mmu_lock); |
321 | /* | |
322 | * The count increase must become visible at unlock time as no | |
323 | * spte can be established without taking the mmu_lock and | |
324 | * count is also read inside the mmu_lock critical section. | |
325 | */ | |
326 | kvm->mmu_notifier_count++; | |
327 | for (; start < end; start += PAGE_SIZE) | |
328 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
a4ee1ca4 | 329 | need_tlb_flush |= kvm->tlbs_dirty; |
e930bffe | 330 | spin_unlock(&kvm->mmu_lock); |
bc6678a3 | 331 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
332 | |
333 | /* we've to flush the tlb before the pages can be freed */ | |
334 | if (need_tlb_flush) | |
335 | kvm_flush_remote_tlbs(kvm); | |
336 | } | |
337 | ||
338 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
339 | struct mm_struct *mm, | |
340 | unsigned long start, | |
341 | unsigned long end) | |
342 | { | |
343 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
344 | ||
345 | spin_lock(&kvm->mmu_lock); | |
346 | /* | |
347 | * This sequence increase will notify the kvm page fault that | |
348 | * the page that is going to be mapped in the spte could have | |
349 | * been freed. | |
350 | */ | |
351 | kvm->mmu_notifier_seq++; | |
352 | /* | |
353 | * The above sequence increase must be visible before the | |
354 | * below count decrease but both values are read by the kvm | |
355 | * page fault under mmu_lock spinlock so we don't need to add | |
356 | * a smb_wmb() here in between the two. | |
357 | */ | |
358 | kvm->mmu_notifier_count--; | |
359 | spin_unlock(&kvm->mmu_lock); | |
360 | ||
361 | BUG_ON(kvm->mmu_notifier_count < 0); | |
362 | } | |
363 | ||
364 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
365 | struct mm_struct *mm, | |
366 | unsigned long address) | |
367 | { | |
368 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 369 | int young, idx; |
e930bffe | 370 | |
bc6678a3 | 371 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
372 | spin_lock(&kvm->mmu_lock); |
373 | young = kvm_age_hva(kvm, address); | |
374 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 375 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
376 | |
377 | if (young) | |
378 | kvm_flush_remote_tlbs(kvm); | |
379 | ||
380 | return young; | |
381 | } | |
382 | ||
85db06e5 MT |
383 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
384 | struct mm_struct *mm) | |
385 | { | |
386 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
387 | int idx; |
388 | ||
389 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 390 | kvm_arch_flush_shadow(kvm); |
eda2beda | 391 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
392 | } |
393 | ||
e930bffe AA |
394 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
395 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
396 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
397 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
398 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
3da0dd43 | 399 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 400 | .release = kvm_mmu_notifier_release, |
e930bffe | 401 | }; |
4c07b0a4 AK |
402 | |
403 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
404 | { | |
405 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
406 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
407 | } | |
408 | ||
409 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
410 | ||
411 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
412 | { | |
413 | return 0; | |
414 | } | |
415 | ||
e930bffe AA |
416 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
417 | ||
f17abe9a | 418 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 419 | { |
d89f5eff JK |
420 | int r, i; |
421 | struct kvm *kvm = kvm_arch_alloc_vm(); | |
6aa8b732 | 422 | |
d89f5eff JK |
423 | if (!kvm) |
424 | return ERR_PTR(-ENOMEM); | |
425 | ||
426 | r = kvm_arch_init_vm(kvm); | |
427 | if (r) | |
428 | goto out_err_nodisable; | |
10474ae8 AG |
429 | |
430 | r = hardware_enable_all(); | |
431 | if (r) | |
432 | goto out_err_nodisable; | |
433 | ||
75858a84 AK |
434 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
435 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 436 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 437 | #endif |
6aa8b732 | 438 | |
46a26bf5 MT |
439 | r = -ENOMEM; |
440 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
441 | if (!kvm->memslots) | |
57e7fbee | 442 | goto out_err_nosrcu; |
bc6678a3 | 443 | if (init_srcu_struct(&kvm->srcu)) |
57e7fbee | 444 | goto out_err_nosrcu; |
e93f8a0f MT |
445 | for (i = 0; i < KVM_NR_BUSES; i++) { |
446 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
447 | GFP_KERNEL); | |
57e7fbee | 448 | if (!kvm->buses[i]) |
e93f8a0f | 449 | goto out_err; |
e93f8a0f | 450 | } |
46a26bf5 | 451 | |
4c07b0a4 | 452 | r = kvm_init_mmu_notifier(kvm); |
57e7fbee | 453 | if (r) |
283d0c65 | 454 | goto out_err; |
e930bffe | 455 | |
6d4e4c4f AK |
456 | kvm->mm = current->mm; |
457 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 458 | spin_lock_init(&kvm->mmu_lock); |
70e335e1 | 459 | raw_spin_lock_init(&kvm->requests_lock); |
d34e6b17 | 460 | kvm_eventfd_init(kvm); |
11ec2804 | 461 | mutex_init(&kvm->lock); |
60eead79 | 462 | mutex_init(&kvm->irq_lock); |
79fac95e | 463 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 464 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
465 | spin_lock(&kvm_lock); |
466 | list_add(&kvm->vm_list, &vm_list); | |
467 | spin_unlock(&kvm_lock); | |
d89f5eff | 468 | |
f17abe9a | 469 | return kvm; |
10474ae8 AG |
470 | |
471 | out_err: | |
57e7fbee JK |
472 | cleanup_srcu_struct(&kvm->srcu); |
473 | out_err_nosrcu: | |
10474ae8 AG |
474 | hardware_disable_all(); |
475 | out_err_nodisable: | |
e93f8a0f MT |
476 | for (i = 0; i < KVM_NR_BUSES; i++) |
477 | kfree(kvm->buses[i]); | |
46a26bf5 | 478 | kfree(kvm->memslots); |
d89f5eff | 479 | kvm_arch_free_vm(kvm); |
10474ae8 | 480 | return ERR_PTR(r); |
f17abe9a AK |
481 | } |
482 | ||
a36a57b1 TY |
483 | static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot) |
484 | { | |
485 | if (!memslot->dirty_bitmap) | |
486 | return; | |
487 | ||
6f9e5c17 TY |
488 | if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE) |
489 | vfree(memslot->dirty_bitmap_head); | |
490 | else | |
491 | kfree(memslot->dirty_bitmap_head); | |
492 | ||
a36a57b1 | 493 | memslot->dirty_bitmap = NULL; |
515a0127 | 494 | memslot->dirty_bitmap_head = NULL; |
a36a57b1 TY |
495 | } |
496 | ||
6aa8b732 AK |
497 | /* |
498 | * Free any memory in @free but not in @dont. | |
499 | */ | |
500 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
501 | struct kvm_memory_slot *dont) | |
502 | { | |
ec04b260 JR |
503 | int i; |
504 | ||
290fc38d IE |
505 | if (!dont || free->rmap != dont->rmap) |
506 | vfree(free->rmap); | |
6aa8b732 AK |
507 | |
508 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
a36a57b1 | 509 | kvm_destroy_dirty_bitmap(free); |
6aa8b732 | 510 | |
ec04b260 JR |
511 | |
512 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
513 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
514 | vfree(free->lpage_info[i]); | |
515 | free->lpage_info[i] = NULL; | |
516 | } | |
517 | } | |
05da4558 | 518 | |
6aa8b732 | 519 | free->npages = 0; |
8d4e1288 | 520 | free->rmap = NULL; |
6aa8b732 AK |
521 | } |
522 | ||
d19a9cd2 | 523 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
524 | { |
525 | int i; | |
46a26bf5 MT |
526 | struct kvm_memslots *slots = kvm->memslots; |
527 | ||
528 | for (i = 0; i < slots->nmemslots; ++i) | |
529 | kvm_free_physmem_slot(&slots->memslots[i], NULL); | |
6aa8b732 | 530 | |
46a26bf5 | 531 | kfree(kvm->memslots); |
6aa8b732 AK |
532 | } |
533 | ||
f17abe9a AK |
534 | static void kvm_destroy_vm(struct kvm *kvm) |
535 | { | |
e93f8a0f | 536 | int i; |
6d4e4c4f AK |
537 | struct mm_struct *mm = kvm->mm; |
538 | ||
ad8ba2cd | 539 | kvm_arch_sync_events(kvm); |
133de902 AK |
540 | spin_lock(&kvm_lock); |
541 | list_del(&kvm->vm_list); | |
542 | spin_unlock(&kvm_lock); | |
399ec807 | 543 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
544 | for (i = 0; i < KVM_NR_BUSES; i++) |
545 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 546 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
547 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
548 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
549 | #else |
550 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 551 | #endif |
d19a9cd2 | 552 | kvm_arch_destroy_vm(kvm); |
d89f5eff JK |
553 | kvm_free_physmem(kvm); |
554 | cleanup_srcu_struct(&kvm->srcu); | |
555 | kvm_arch_free_vm(kvm); | |
10474ae8 | 556 | hardware_disable_all(); |
6d4e4c4f | 557 | mmdrop(mm); |
f17abe9a AK |
558 | } |
559 | ||
d39f13b0 IE |
560 | void kvm_get_kvm(struct kvm *kvm) |
561 | { | |
562 | atomic_inc(&kvm->users_count); | |
563 | } | |
564 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
565 | ||
566 | void kvm_put_kvm(struct kvm *kvm) | |
567 | { | |
568 | if (atomic_dec_and_test(&kvm->users_count)) | |
569 | kvm_destroy_vm(kvm); | |
570 | } | |
571 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
572 | ||
573 | ||
f17abe9a AK |
574 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
575 | { | |
576 | struct kvm *kvm = filp->private_data; | |
577 | ||
721eecbf GH |
578 | kvm_irqfd_release(kvm); |
579 | ||
d39f13b0 | 580 | kvm_put_kvm(kvm); |
6aa8b732 AK |
581 | return 0; |
582 | } | |
583 | ||
515a0127 TY |
584 | /* |
585 | * Allocation size is twice as large as the actual dirty bitmap size. | |
586 | * This makes it possible to do double buffering: see x86's | |
587 | * kvm_vm_ioctl_get_dirty_log(). | |
588 | */ | |
a36a57b1 TY |
589 | static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot) |
590 | { | |
515a0127 | 591 | unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot); |
a36a57b1 | 592 | |
6f9e5c17 TY |
593 | if (dirty_bytes > PAGE_SIZE) |
594 | memslot->dirty_bitmap = vzalloc(dirty_bytes); | |
595 | else | |
596 | memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL); | |
597 | ||
a36a57b1 TY |
598 | if (!memslot->dirty_bitmap) |
599 | return -ENOMEM; | |
600 | ||
515a0127 | 601 | memslot->dirty_bitmap_head = memslot->dirty_bitmap; |
a36a57b1 TY |
602 | return 0; |
603 | } | |
604 | ||
6aa8b732 AK |
605 | /* |
606 | * Allocate some memory and give it an address in the guest physical address | |
607 | * space. | |
608 | * | |
609 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 610 | * |
10589a46 | 611 | * Must be called holding mmap_sem for write. |
6aa8b732 | 612 | */ |
f78e0e2e SY |
613 | int __kvm_set_memory_region(struct kvm *kvm, |
614 | struct kvm_userspace_memory_region *mem, | |
615 | int user_alloc) | |
6aa8b732 | 616 | { |
bc6678a3 | 617 | int r, flush_shadow = 0; |
6aa8b732 | 618 | gfn_t base_gfn; |
28bcb112 HC |
619 | unsigned long npages; |
620 | unsigned long i; | |
6aa8b732 AK |
621 | struct kvm_memory_slot *memslot; |
622 | struct kvm_memory_slot old, new; | |
bc6678a3 | 623 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
624 | |
625 | r = -EINVAL; | |
626 | /* General sanity checks */ | |
627 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
628 | goto out; | |
629 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
630 | goto out; | |
e7cacd40 | 631 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 632 | goto out; |
e0d62c7f | 633 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
634 | goto out; |
635 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
636 | goto out; | |
637 | ||
46a26bf5 | 638 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
639 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
640 | npages = mem->memory_size >> PAGE_SHIFT; | |
641 | ||
660c22c4 TY |
642 | r = -EINVAL; |
643 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
644 | goto out; | |
645 | ||
6aa8b732 AK |
646 | if (!npages) |
647 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
648 | ||
6aa8b732 AK |
649 | new = old = *memslot; |
650 | ||
e36d96f7 | 651 | new.id = mem->slot; |
6aa8b732 AK |
652 | new.base_gfn = base_gfn; |
653 | new.npages = npages; | |
654 | new.flags = mem->flags; | |
655 | ||
656 | /* Disallow changing a memory slot's size. */ | |
657 | r = -EINVAL; | |
658 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 659 | goto out_free; |
6aa8b732 AK |
660 | |
661 | /* Check for overlaps */ | |
662 | r = -EEXIST; | |
663 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 664 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 665 | |
4cd481f6 | 666 | if (s == memslot || !s->npages) |
6aa8b732 AK |
667 | continue; |
668 | if (!((base_gfn + npages <= s->base_gfn) || | |
669 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 670 | goto out_free; |
6aa8b732 | 671 | } |
6aa8b732 | 672 | |
6aa8b732 AK |
673 | /* Free page dirty bitmap if unneeded */ |
674 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 675 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
676 | |
677 | r = -ENOMEM; | |
678 | ||
679 | /* Allocate if a slot is being created */ | |
eff0114a | 680 | #ifndef CONFIG_S390 |
8d4e1288 | 681 | if (npages && !new.rmap) { |
26535037 | 682 | new.rmap = vzalloc(npages * sizeof(*new.rmap)); |
290fc38d IE |
683 | |
684 | if (!new.rmap) | |
f78e0e2e | 685 | goto out_free; |
290fc38d | 686 | |
80b14b5b | 687 | new.user_alloc = user_alloc; |
bc6678a3 | 688 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 689 | } |
ec04b260 JR |
690 | if (!npages) |
691 | goto skip_lpage; | |
05da4558 | 692 | |
ec04b260 | 693 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
694 | unsigned long ugfn; |
695 | unsigned long j; | |
696 | int lpages; | |
ec04b260 | 697 | int level = i + 2; |
05da4558 | 698 | |
ec04b260 JR |
699 | /* Avoid unused variable warning if no large pages */ |
700 | (void)level; | |
701 | ||
702 | if (new.lpage_info[i]) | |
703 | continue; | |
704 | ||
82855413 JR |
705 | lpages = 1 + ((base_gfn + npages - 1) |
706 | >> KVM_HPAGE_GFN_SHIFT(level)); | |
707 | lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level); | |
ec04b260 | 708 | |
26535037 | 709 | new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i])); |
ec04b260 JR |
710 | |
711 | if (!new.lpage_info[i]) | |
05da4558 MT |
712 | goto out_free; |
713 | ||
82855413 | 714 | if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 715 | new.lpage_info[i][0].write_count = 1; |
82855413 | 716 | if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 717 | new.lpage_info[i][lpages - 1].write_count = 1; |
ac04527f AK |
718 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
719 | /* | |
720 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
721 | * other, or if explicitly asked to, disable large page |
722 | * support for this slot | |
ac04527f | 723 | */ |
ec04b260 | 724 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 725 | !largepages_enabled) |
ec04b260 JR |
726 | for (j = 0; j < lpages; ++j) |
727 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 728 | } |
6aa8b732 | 729 | |
ec04b260 JR |
730 | skip_lpage: |
731 | ||
6aa8b732 AK |
732 | /* Allocate page dirty bitmap if needed */ |
733 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
a36a57b1 | 734 | if (kvm_create_dirty_bitmap(&new) < 0) |
f78e0e2e | 735 | goto out_free; |
bc6678a3 | 736 | /* destroy any largepage mappings for dirty tracking */ |
e244584f | 737 | if (old.npages) |
bc6678a3 | 738 | flush_shadow = 1; |
6aa8b732 | 739 | } |
3eea8437 CB |
740 | #else /* not defined CONFIG_S390 */ |
741 | new.user_alloc = user_alloc; | |
742 | if (user_alloc) | |
743 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 744 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 745 | |
bc6678a3 MT |
746 | if (!npages) { |
747 | r = -ENOMEM; | |
748 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
749 | if (!slots) | |
750 | goto out_free; | |
751 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
752 | if (mem->slot >= slots->nmemslots) | |
753 | slots->nmemslots = mem->slot + 1; | |
49c7754c | 754 | slots->generation++; |
bc6678a3 MT |
755 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; |
756 | ||
757 | old_memslots = kvm->memslots; | |
758 | rcu_assign_pointer(kvm->memslots, slots); | |
759 | synchronize_srcu_expedited(&kvm->srcu); | |
760 | /* From this point no new shadow pages pointing to a deleted | |
761 | * memslot will be created. | |
762 | * | |
763 | * validation of sp->gfn happens in: | |
764 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
765 | * - kvm_is_visible_gfn (mmu_check_roots) | |
766 | */ | |
34d4cb8f | 767 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
768 | kfree(old_memslots); |
769 | } | |
34d4cb8f | 770 | |
f7784b8e MT |
771 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
772 | if (r) | |
773 | goto out_free; | |
774 | ||
bc6678a3 MT |
775 | /* map the pages in iommu page table */ |
776 | if (npages) { | |
777 | r = kvm_iommu_map_pages(kvm, &new); | |
778 | if (r) | |
779 | goto out_free; | |
780 | } | |
604b38ac | 781 | |
bc6678a3 MT |
782 | r = -ENOMEM; |
783 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
784 | if (!slots) | |
785 | goto out_free; | |
786 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
787 | if (mem->slot >= slots->nmemslots) | |
788 | slots->nmemslots = mem->slot + 1; | |
49c7754c | 789 | slots->generation++; |
bc6678a3 MT |
790 | |
791 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
792 | if (!npages) { | |
793 | new.rmap = NULL; | |
794 | new.dirty_bitmap = NULL; | |
795 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
796 | new.lpage_info[i] = NULL; | |
797 | } | |
798 | ||
799 | slots->memslots[mem->slot] = new; | |
800 | old_memslots = kvm->memslots; | |
801 | rcu_assign_pointer(kvm->memslots, slots); | |
802 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 803 | |
f7784b8e | 804 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 805 | |
bc6678a3 MT |
806 | kvm_free_physmem_slot(&old, &new); |
807 | kfree(old_memslots); | |
808 | ||
809 | if (flush_shadow) | |
810 | kvm_arch_flush_shadow(kvm); | |
811 | ||
6aa8b732 AK |
812 | return 0; |
813 | ||
f78e0e2e | 814 | out_free: |
6aa8b732 AK |
815 | kvm_free_physmem_slot(&new, &old); |
816 | out: | |
817 | return r; | |
210c7c4d IE |
818 | |
819 | } | |
f78e0e2e SY |
820 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
821 | ||
822 | int kvm_set_memory_region(struct kvm *kvm, | |
823 | struct kvm_userspace_memory_region *mem, | |
824 | int user_alloc) | |
825 | { | |
826 | int r; | |
827 | ||
79fac95e | 828 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 829 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 830 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
831 | return r; |
832 | } | |
210c7c4d IE |
833 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
834 | ||
1fe779f8 CO |
835 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
836 | struct | |
837 | kvm_userspace_memory_region *mem, | |
838 | int user_alloc) | |
210c7c4d | 839 | { |
e0d62c7f IE |
840 | if (mem->slot >= KVM_MEMORY_SLOTS) |
841 | return -EINVAL; | |
210c7c4d | 842 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
843 | } |
844 | ||
5bb064dc ZX |
845 | int kvm_get_dirty_log(struct kvm *kvm, |
846 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
847 | { |
848 | struct kvm_memory_slot *memslot; | |
849 | int r, i; | |
87bf6e7d | 850 | unsigned long n; |
6aa8b732 AK |
851 | unsigned long any = 0; |
852 | ||
6aa8b732 AK |
853 | r = -EINVAL; |
854 | if (log->slot >= KVM_MEMORY_SLOTS) | |
855 | goto out; | |
856 | ||
46a26bf5 | 857 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
858 | r = -ENOENT; |
859 | if (!memslot->dirty_bitmap) | |
860 | goto out; | |
861 | ||
87bf6e7d | 862 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 863 | |
cd1a4a98 | 864 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
865 | any = memslot->dirty_bitmap[i]; |
866 | ||
867 | r = -EFAULT; | |
868 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
869 | goto out; | |
870 | ||
5bb064dc ZX |
871 | if (any) |
872 | *is_dirty = 1; | |
6aa8b732 AK |
873 | |
874 | r = 0; | |
6aa8b732 | 875 | out: |
6aa8b732 AK |
876 | return r; |
877 | } | |
878 | ||
54dee993 MT |
879 | void kvm_disable_largepages(void) |
880 | { | |
881 | largepages_enabled = false; | |
882 | } | |
883 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
884 | ||
cea7bb21 IE |
885 | int is_error_page(struct page *page) |
886 | { | |
edba23e5 | 887 | return page == bad_page || page == hwpoison_page || page == fault_page; |
cea7bb21 IE |
888 | } |
889 | EXPORT_SYMBOL_GPL(is_error_page); | |
890 | ||
35149e21 AL |
891 | int is_error_pfn(pfn_t pfn) |
892 | { | |
edba23e5 | 893 | return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn; |
35149e21 AL |
894 | } |
895 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
896 | ||
bf998156 HY |
897 | int is_hwpoison_pfn(pfn_t pfn) |
898 | { | |
899 | return pfn == hwpoison_pfn; | |
900 | } | |
901 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
902 | ||
edba23e5 GN |
903 | int is_fault_pfn(pfn_t pfn) |
904 | { | |
905 | return pfn == fault_pfn; | |
906 | } | |
907 | EXPORT_SYMBOL_GPL(is_fault_pfn); | |
908 | ||
f9d46eb0 IE |
909 | static inline unsigned long bad_hva(void) |
910 | { | |
911 | return PAGE_OFFSET; | |
912 | } | |
913 | ||
914 | int kvm_is_error_hva(unsigned long addr) | |
915 | { | |
916 | return addr == bad_hva(); | |
917 | } | |
918 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
919 | ||
49c7754c GN |
920 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots, |
921 | gfn_t gfn) | |
6aa8b732 AK |
922 | { |
923 | int i; | |
924 | ||
46a26bf5 MT |
925 | for (i = 0; i < slots->nmemslots; ++i) { |
926 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
927 | |
928 | if (gfn >= memslot->base_gfn | |
929 | && gfn < memslot->base_gfn + memslot->npages) | |
930 | return memslot; | |
931 | } | |
8b6d44c7 | 932 | return NULL; |
6aa8b732 | 933 | } |
49c7754c GN |
934 | |
935 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
936 | { | |
937 | return __gfn_to_memslot(kvm_memslots(kvm), gfn); | |
938 | } | |
a1f4d395 | 939 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 940 | |
e0d62c7f IE |
941 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
942 | { | |
943 | int i; | |
90d83dc3 | 944 | struct kvm_memslots *slots = kvm_memslots(kvm); |
e0d62c7f | 945 | |
e0d62c7f | 946 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 947 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 948 | |
bc6678a3 MT |
949 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
950 | continue; | |
951 | ||
e0d62c7f IE |
952 | if (gfn >= memslot->base_gfn |
953 | && gfn < memslot->base_gfn + memslot->npages) | |
954 | return 1; | |
955 | } | |
956 | return 0; | |
957 | } | |
958 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
959 | ||
8f0b1ab6 JR |
960 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
961 | { | |
962 | struct vm_area_struct *vma; | |
963 | unsigned long addr, size; | |
964 | ||
965 | size = PAGE_SIZE; | |
966 | ||
967 | addr = gfn_to_hva(kvm, gfn); | |
968 | if (kvm_is_error_hva(addr)) | |
969 | return PAGE_SIZE; | |
970 | ||
971 | down_read(¤t->mm->mmap_sem); | |
972 | vma = find_vma(current->mm, addr); | |
973 | if (!vma) | |
974 | goto out; | |
975 | ||
976 | size = vma_kernel_pagesize(vma); | |
977 | ||
978 | out: | |
979 | up_read(¤t->mm->mmap_sem); | |
980 | ||
981 | return size; | |
982 | } | |
983 | ||
bc6678a3 MT |
984 | int memslot_id(struct kvm *kvm, gfn_t gfn) |
985 | { | |
986 | int i; | |
90d83dc3 | 987 | struct kvm_memslots *slots = kvm_memslots(kvm); |
bc6678a3 MT |
988 | struct kvm_memory_slot *memslot = NULL; |
989 | ||
bc6678a3 MT |
990 | for (i = 0; i < slots->nmemslots; ++i) { |
991 | memslot = &slots->memslots[i]; | |
992 | ||
993 | if (gfn >= memslot->base_gfn | |
994 | && gfn < memslot->base_gfn + memslot->npages) | |
995 | break; | |
996 | } | |
997 | ||
998 | return memslot - slots->memslots; | |
999 | } | |
1000 | ||
49c7754c | 1001 | static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn, |
48987781 | 1002 | gfn_t *nr_pages) |
539cb660 | 1003 | { |
bc6678a3 | 1004 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 1005 | return bad_hva(); |
48987781 XG |
1006 | |
1007 | if (nr_pages) | |
1008 | *nr_pages = slot->npages - (gfn - slot->base_gfn); | |
1009 | ||
f5c98031 | 1010 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 1011 | } |
48987781 XG |
1012 | |
1013 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) | |
1014 | { | |
49c7754c | 1015 | return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL); |
48987781 | 1016 | } |
0d150298 | 1017 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1018 | |
8030089f GN |
1019 | static pfn_t get_fault_pfn(void) |
1020 | { | |
1021 | get_page(fault_page); | |
1022 | return fault_pfn; | |
1023 | } | |
1024 | ||
af585b92 | 1025 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic, |
612819c3 | 1026 | bool *async, bool write_fault, bool *writable) |
954bbbc2 | 1027 | { |
8d4e1288 | 1028 | struct page *page[1]; |
af585b92 | 1029 | int npages = 0; |
2e2e3738 | 1030 | pfn_t pfn; |
954bbbc2 | 1031 | |
af585b92 GN |
1032 | /* we can do it either atomically or asynchronously, not both */ |
1033 | BUG_ON(atomic && async); | |
1034 | ||
612819c3 MT |
1035 | BUG_ON(!write_fault && !writable); |
1036 | ||
1037 | if (writable) | |
1038 | *writable = true; | |
1039 | ||
af585b92 | 1040 | if (atomic || async) |
887c08ac | 1041 | npages = __get_user_pages_fast(addr, 1, 1, page); |
af585b92 GN |
1042 | |
1043 | if (unlikely(npages != 1) && !atomic) { | |
887c08ac | 1044 | might_sleep(); |
612819c3 MT |
1045 | |
1046 | if (writable) | |
1047 | *writable = write_fault; | |
1048 | ||
1049 | npages = get_user_pages_fast(addr, 1, write_fault, page); | |
1050 | ||
1051 | /* map read fault as writable if possible */ | |
1052 | if (unlikely(!write_fault) && npages == 1) { | |
1053 | struct page *wpage[1]; | |
1054 | ||
1055 | npages = __get_user_pages_fast(addr, 1, 1, wpage); | |
1056 | if (npages == 1) { | |
1057 | *writable = true; | |
1058 | put_page(page[0]); | |
1059 | page[0] = wpage[0]; | |
1060 | } | |
1061 | npages = 1; | |
1062 | } | |
887c08ac | 1063 | } |
539cb660 | 1064 | |
2e2e3738 AL |
1065 | if (unlikely(npages != 1)) { |
1066 | struct vm_area_struct *vma; | |
1067 | ||
887c08ac | 1068 | if (atomic) |
8030089f | 1069 | return get_fault_pfn(); |
887c08ac | 1070 | |
bbeb3406 | 1071 | down_read(¤t->mm->mmap_sem); |
bf998156 | 1072 | if (is_hwpoison_address(addr)) { |
bbeb3406 | 1073 | up_read(¤t->mm->mmap_sem); |
bf998156 HY |
1074 | get_page(hwpoison_page); |
1075 | return page_to_pfn(hwpoison_page); | |
1076 | } | |
1077 | ||
8030089f | 1078 | vma = find_vma_intersection(current->mm, addr, addr+1); |
4c2155ce | 1079 | |
8030089f GN |
1080 | if (vma == NULL) |
1081 | pfn = get_fault_pfn(); | |
1082 | else if ((vma->vm_flags & VM_PFNMAP)) { | |
1083 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + | |
1084 | vma->vm_pgoff; | |
1085 | BUG_ON(!kvm_is_mmio_pfn(pfn)); | |
1086 | } else { | |
1087 | if (async && (vma->vm_flags & VM_WRITE)) | |
af585b92 | 1088 | *async = true; |
8030089f | 1089 | pfn = get_fault_pfn(); |
2e2e3738 | 1090 | } |
4c2155ce | 1091 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1092 | } else |
1093 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1094 | |
2e2e3738 | 1095 | return pfn; |
35149e21 AL |
1096 | } |
1097 | ||
887c08ac XG |
1098 | pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr) |
1099 | { | |
612819c3 | 1100 | return hva_to_pfn(kvm, addr, true, NULL, true, NULL); |
887c08ac XG |
1101 | } |
1102 | EXPORT_SYMBOL_GPL(hva_to_pfn_atomic); | |
1103 | ||
612819c3 MT |
1104 | static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async, |
1105 | bool write_fault, bool *writable) | |
506f0d6f MT |
1106 | { |
1107 | unsigned long addr; | |
1108 | ||
af585b92 GN |
1109 | if (async) |
1110 | *async = false; | |
1111 | ||
506f0d6f MT |
1112 | addr = gfn_to_hva(kvm, gfn); |
1113 | if (kvm_is_error_hva(addr)) { | |
1114 | get_page(bad_page); | |
1115 | return page_to_pfn(bad_page); | |
1116 | } | |
1117 | ||
612819c3 | 1118 | return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable); |
365fb3fd XG |
1119 | } |
1120 | ||
1121 | pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn) | |
1122 | { | |
612819c3 | 1123 | return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL); |
365fb3fd XG |
1124 | } |
1125 | EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic); | |
1126 | ||
612819c3 MT |
1127 | pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async, |
1128 | bool write_fault, bool *writable) | |
af585b92 | 1129 | { |
612819c3 | 1130 | return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable); |
af585b92 GN |
1131 | } |
1132 | EXPORT_SYMBOL_GPL(gfn_to_pfn_async); | |
1133 | ||
365fb3fd XG |
1134 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
1135 | { | |
612819c3 | 1136 | return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL); |
506f0d6f | 1137 | } |
35149e21 AL |
1138 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
1139 | ||
612819c3 MT |
1140 | pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, |
1141 | bool *writable) | |
1142 | { | |
1143 | return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable); | |
1144 | } | |
1145 | EXPORT_SYMBOL_GPL(gfn_to_pfn_prot); | |
1146 | ||
506f0d6f MT |
1147 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
1148 | struct kvm_memory_slot *slot, gfn_t gfn) | |
1149 | { | |
1150 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
612819c3 | 1151 | return hva_to_pfn(kvm, addr, false, NULL, true, NULL); |
506f0d6f MT |
1152 | } |
1153 | ||
48987781 XG |
1154 | int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages, |
1155 | int nr_pages) | |
1156 | { | |
1157 | unsigned long addr; | |
1158 | gfn_t entry; | |
1159 | ||
49c7754c | 1160 | addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry); |
48987781 XG |
1161 | if (kvm_is_error_hva(addr)) |
1162 | return -1; | |
1163 | ||
1164 | if (entry < nr_pages) | |
1165 | return 0; | |
1166 | ||
1167 | return __get_user_pages_fast(addr, nr_pages, 1, pages); | |
1168 | } | |
1169 | EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic); | |
1170 | ||
35149e21 AL |
1171 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1172 | { | |
2e2e3738 AL |
1173 | pfn_t pfn; |
1174 | ||
1175 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1176 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1177 | return pfn_to_page(pfn); |
1178 | ||
c77fb9dc | 1179 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1180 | |
1181 | get_page(bad_page); | |
1182 | return bad_page; | |
954bbbc2 | 1183 | } |
aab61cc0 | 1184 | |
954bbbc2 AK |
1185 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1186 | ||
b4231d61 IE |
1187 | void kvm_release_page_clean(struct page *page) |
1188 | { | |
35149e21 | 1189 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1190 | } |
1191 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1192 | ||
35149e21 AL |
1193 | void kvm_release_pfn_clean(pfn_t pfn) |
1194 | { | |
c77fb9dc | 1195 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1196 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1197 | } |
1198 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1199 | ||
b4231d61 | 1200 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1201 | { |
35149e21 AL |
1202 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1203 | } | |
1204 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1205 | ||
1206 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1207 | { | |
1208 | kvm_set_pfn_dirty(pfn); | |
1209 | kvm_release_pfn_clean(pfn); | |
1210 | } | |
1211 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1212 | ||
1213 | void kvm_set_page_dirty(struct page *page) | |
1214 | { | |
1215 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1216 | } | |
1217 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1218 | ||
1219 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1220 | { | |
c77fb9dc | 1221 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1222 | struct page *page = pfn_to_page(pfn); |
1223 | if (!PageReserved(page)) | |
1224 | SetPageDirty(page); | |
1225 | } | |
8a7ae055 | 1226 | } |
35149e21 AL |
1227 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1228 | ||
1229 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1230 | { | |
c77fb9dc | 1231 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1232 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1233 | } |
1234 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1235 | ||
1236 | void kvm_get_pfn(pfn_t pfn) | |
1237 | { | |
c77fb9dc | 1238 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1239 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1240 | } |
1241 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1242 | |
195aefde IE |
1243 | static int next_segment(unsigned long len, int offset) |
1244 | { | |
1245 | if (len > PAGE_SIZE - offset) | |
1246 | return PAGE_SIZE - offset; | |
1247 | else | |
1248 | return len; | |
1249 | } | |
1250 | ||
1251 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1252 | int len) | |
1253 | { | |
e0506bcb IE |
1254 | int r; |
1255 | unsigned long addr; | |
195aefde | 1256 | |
e0506bcb IE |
1257 | addr = gfn_to_hva(kvm, gfn); |
1258 | if (kvm_is_error_hva(addr)) | |
1259 | return -EFAULT; | |
1260 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1261 | if (r) | |
195aefde | 1262 | return -EFAULT; |
195aefde IE |
1263 | return 0; |
1264 | } | |
1265 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1266 | ||
1267 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1268 | { | |
1269 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1270 | int seg; | |
1271 | int offset = offset_in_page(gpa); | |
1272 | int ret; | |
1273 | ||
1274 | while ((seg = next_segment(len, offset)) != 0) { | |
1275 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1276 | if (ret < 0) | |
1277 | return ret; | |
1278 | offset = 0; | |
1279 | len -= seg; | |
1280 | data += seg; | |
1281 | ++gfn; | |
1282 | } | |
1283 | return 0; | |
1284 | } | |
1285 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1286 | ||
7ec54588 MT |
1287 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1288 | unsigned long len) | |
1289 | { | |
1290 | int r; | |
1291 | unsigned long addr; | |
1292 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1293 | int offset = offset_in_page(gpa); | |
1294 | ||
1295 | addr = gfn_to_hva(kvm, gfn); | |
1296 | if (kvm_is_error_hva(addr)) | |
1297 | return -EFAULT; | |
0aac03f0 | 1298 | pagefault_disable(); |
7ec54588 | 1299 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1300 | pagefault_enable(); |
7ec54588 MT |
1301 | if (r) |
1302 | return -EFAULT; | |
1303 | return 0; | |
1304 | } | |
1305 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1306 | ||
195aefde IE |
1307 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1308 | int offset, int len) | |
1309 | { | |
e0506bcb IE |
1310 | int r; |
1311 | unsigned long addr; | |
195aefde | 1312 | |
e0506bcb IE |
1313 | addr = gfn_to_hva(kvm, gfn); |
1314 | if (kvm_is_error_hva(addr)) | |
1315 | return -EFAULT; | |
1316 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1317 | if (r) | |
195aefde | 1318 | return -EFAULT; |
195aefde IE |
1319 | mark_page_dirty(kvm, gfn); |
1320 | return 0; | |
1321 | } | |
1322 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1323 | ||
1324 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1325 | unsigned long len) | |
1326 | { | |
1327 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1328 | int seg; | |
1329 | int offset = offset_in_page(gpa); | |
1330 | int ret; | |
1331 | ||
1332 | while ((seg = next_segment(len, offset)) != 0) { | |
1333 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1334 | if (ret < 0) | |
1335 | return ret; | |
1336 | offset = 0; | |
1337 | len -= seg; | |
1338 | data += seg; | |
1339 | ++gfn; | |
1340 | } | |
1341 | return 0; | |
1342 | } | |
1343 | ||
49c7754c GN |
1344 | int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1345 | gpa_t gpa) | |
1346 | { | |
1347 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1348 | int offset = offset_in_page(gpa); | |
1349 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1350 | ||
1351 | ghc->gpa = gpa; | |
1352 | ghc->generation = slots->generation; | |
1353 | ghc->memslot = __gfn_to_memslot(slots, gfn); | |
1354 | ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL); | |
1355 | if (!kvm_is_error_hva(ghc->hva)) | |
1356 | ghc->hva += offset; | |
1357 | else | |
1358 | return -EFAULT; | |
1359 | ||
1360 | return 0; | |
1361 | } | |
1362 | EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init); | |
1363 | ||
1364 | int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, | |
1365 | void *data, unsigned long len) | |
1366 | { | |
1367 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1368 | int r; | |
1369 | ||
1370 | if (slots->generation != ghc->generation) | |
1371 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1372 | ||
1373 | if (kvm_is_error_hva(ghc->hva)) | |
1374 | return -EFAULT; | |
1375 | ||
1376 | r = copy_to_user((void __user *)ghc->hva, data, len); | |
1377 | if (r) | |
1378 | return -EFAULT; | |
1379 | mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT); | |
1380 | ||
1381 | return 0; | |
1382 | } | |
1383 | EXPORT_SYMBOL_GPL(kvm_write_guest_cached); | |
1384 | ||
195aefde IE |
1385 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) |
1386 | { | |
3bcc8a8c HC |
1387 | return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page, |
1388 | offset, len); | |
195aefde IE |
1389 | } |
1390 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1391 | ||
1392 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1393 | { | |
1394 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1395 | int seg; | |
1396 | int offset = offset_in_page(gpa); | |
1397 | int ret; | |
1398 | ||
1399 | while ((seg = next_segment(len, offset)) != 0) { | |
1400 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1401 | if (ret < 0) | |
1402 | return ret; | |
1403 | offset = 0; | |
1404 | len -= seg; | |
1405 | ++gfn; | |
1406 | } | |
1407 | return 0; | |
1408 | } | |
1409 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1410 | ||
49c7754c GN |
1411 | void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, |
1412 | gfn_t gfn) | |
6aa8b732 | 1413 | { |
7e9d619d RR |
1414 | if (memslot && memslot->dirty_bitmap) { |
1415 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1416 | |
d1476937 | 1417 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); |
6aa8b732 AK |
1418 | } |
1419 | } | |
1420 | ||
49c7754c GN |
1421 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1422 | { | |
1423 | struct kvm_memory_slot *memslot; | |
1424 | ||
1425 | memslot = gfn_to_memslot(kvm, gfn); | |
1426 | mark_page_dirty_in_slot(kvm, memslot, gfn); | |
1427 | } | |
1428 | ||
b6958ce4 ED |
1429 | /* |
1430 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1431 | */ | |
8776e519 | 1432 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1433 | { |
e5c239cf MT |
1434 | DEFINE_WAIT(wait); |
1435 | ||
1436 | for (;;) { | |
1437 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1438 | ||
a1b37100 | 1439 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1440 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1441 | break; |
d7690175 | 1442 | } |
09cec754 GN |
1443 | if (kvm_cpu_has_pending_timer(vcpu)) |
1444 | break; | |
e5c239cf MT |
1445 | if (signal_pending(current)) |
1446 | break; | |
1447 | ||
b6958ce4 | 1448 | schedule(); |
b6958ce4 | 1449 | } |
d3bef15f | 1450 | |
e5c239cf | 1451 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1452 | } |
1453 | ||
6aa8b732 AK |
1454 | void kvm_resched(struct kvm_vcpu *vcpu) |
1455 | { | |
3fca0365 YD |
1456 | if (!need_resched()) |
1457 | return; | |
6aa8b732 | 1458 | cond_resched(); |
6aa8b732 AK |
1459 | } |
1460 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1461 | ||
d255f4f2 ZE |
1462 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu) |
1463 | { | |
1464 | ktime_t expires; | |
1465 | DEFINE_WAIT(wait); | |
1466 | ||
1467 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1468 | ||
1469 | /* Sleep for 100 us, and hope lock-holder got scheduled */ | |
1470 | expires = ktime_add_ns(ktime_get(), 100000UL); | |
1471 | schedule_hrtimeout(&expires, HRTIMER_MODE_ABS); | |
1472 | ||
1473 | finish_wait(&vcpu->wq, &wait); | |
1474 | } | |
1475 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1476 | ||
e4a533a4 | 1477 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1478 | { |
1479 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1480 | struct page *page; |
1481 | ||
e4a533a4 | 1482 | if (vmf->pgoff == 0) |
039576c0 | 1483 | page = virt_to_page(vcpu->run); |
09566765 | 1484 | #ifdef CONFIG_X86 |
e4a533a4 | 1485 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1486 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1487 | #endif |
1488 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1489 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1490 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1491 | #endif |
039576c0 | 1492 | else |
e4a533a4 | 1493 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1494 | get_page(page); |
e4a533a4 | 1495 | vmf->page = page; |
1496 | return 0; | |
9a2bb7f4 AK |
1497 | } |
1498 | ||
f0f37e2f | 1499 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1500 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1501 | }; |
1502 | ||
1503 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1504 | { | |
1505 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1506 | return 0; | |
1507 | } | |
1508 | ||
bccf2150 AK |
1509 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1510 | { | |
1511 | struct kvm_vcpu *vcpu = filp->private_data; | |
1512 | ||
66c0b394 | 1513 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1514 | return 0; |
1515 | } | |
1516 | ||
3d3aab1b | 1517 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1518 | .release = kvm_vcpu_release, |
1519 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1520 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1521 | .mmap = kvm_vcpu_mmap, |
6038f373 | 1522 | .llseek = noop_llseek, |
bccf2150 AK |
1523 | }; |
1524 | ||
1525 | /* | |
1526 | * Allocates an inode for the vcpu. | |
1527 | */ | |
1528 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1529 | { | |
628ff7c1 | 1530 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1531 | } |
1532 | ||
c5ea7660 AK |
1533 | /* |
1534 | * Creates some virtual cpus. Good luck creating more than one. | |
1535 | */ | |
73880c80 | 1536 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1537 | { |
1538 | int r; | |
988a2cae | 1539 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1540 | |
73880c80 | 1541 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1542 | if (IS_ERR(vcpu)) |
1543 | return PTR_ERR(vcpu); | |
c5ea7660 | 1544 | |
15ad7146 AK |
1545 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1546 | ||
26e5215f AK |
1547 | r = kvm_arch_vcpu_setup(vcpu); |
1548 | if (r) | |
7d8fece6 | 1549 | return r; |
26e5215f | 1550 | |
11ec2804 | 1551 | mutex_lock(&kvm->lock); |
73880c80 GN |
1552 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1553 | r = -EINVAL; | |
e9b11c17 | 1554 | goto vcpu_destroy; |
fb3f0f51 | 1555 | } |
73880c80 | 1556 | |
988a2cae GN |
1557 | kvm_for_each_vcpu(r, v, kvm) |
1558 | if (v->vcpu_id == id) { | |
73880c80 GN |
1559 | r = -EEXIST; |
1560 | goto vcpu_destroy; | |
1561 | } | |
1562 | ||
1563 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1564 | |
fb3f0f51 | 1565 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1566 | kvm_get_kvm(kvm); |
bccf2150 | 1567 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1568 | if (r < 0) { |
1569 | kvm_put_kvm(kvm); | |
1570 | goto vcpu_destroy; | |
1571 | } | |
1572 | ||
1573 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1574 | smp_wmb(); | |
1575 | atomic_inc(&kvm->online_vcpus); | |
1576 | ||
1577 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1578 | if (kvm->bsp_vcpu_id == id) | |
1579 | kvm->bsp_vcpu = vcpu; | |
1580 | #endif | |
1581 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1582 | return r; |
39c3b86e | 1583 | |
e9b11c17 | 1584 | vcpu_destroy: |
7d8fece6 | 1585 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1586 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1587 | return r; |
1588 | } | |
1589 | ||
1961d276 AK |
1590 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1591 | { | |
1592 | if (sigset) { | |
1593 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1594 | vcpu->sigset_active = 1; | |
1595 | vcpu->sigset = *sigset; | |
1596 | } else | |
1597 | vcpu->sigset_active = 0; | |
1598 | return 0; | |
1599 | } | |
1600 | ||
bccf2150 AK |
1601 | static long kvm_vcpu_ioctl(struct file *filp, |
1602 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1603 | { |
bccf2150 | 1604 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1605 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1606 | int r; |
fa3795a7 DH |
1607 | struct kvm_fpu *fpu = NULL; |
1608 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1609 | |
6d4e4c4f AK |
1610 | if (vcpu->kvm->mm != current->mm) |
1611 | return -EIO; | |
2122ff5e AK |
1612 | |
1613 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1614 | /* | |
1615 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1616 | * so vcpu_load() would break it. | |
1617 | */ | |
1618 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1619 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1620 | #endif | |
1621 | ||
1622 | ||
1623 | vcpu_load(vcpu); | |
6aa8b732 | 1624 | switch (ioctl) { |
9a2bb7f4 | 1625 | case KVM_RUN: |
f0fe5108 AK |
1626 | r = -EINVAL; |
1627 | if (arg) | |
1628 | goto out; | |
b6c7a5dc | 1629 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
64be5007 | 1630 | trace_kvm_userspace_exit(vcpu->run->exit_reason, r); |
6aa8b732 | 1631 | break; |
6aa8b732 | 1632 | case KVM_GET_REGS: { |
3e4bb3ac | 1633 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1634 | |
3e4bb3ac XZ |
1635 | r = -ENOMEM; |
1636 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1637 | if (!kvm_regs) | |
6aa8b732 | 1638 | goto out; |
3e4bb3ac XZ |
1639 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1640 | if (r) | |
1641 | goto out_free1; | |
6aa8b732 | 1642 | r = -EFAULT; |
3e4bb3ac XZ |
1643 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1644 | goto out_free1; | |
6aa8b732 | 1645 | r = 0; |
3e4bb3ac XZ |
1646 | out_free1: |
1647 | kfree(kvm_regs); | |
6aa8b732 AK |
1648 | break; |
1649 | } | |
1650 | case KVM_SET_REGS: { | |
3e4bb3ac | 1651 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1652 | |
3e4bb3ac XZ |
1653 | r = -ENOMEM; |
1654 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1655 | if (!kvm_regs) | |
6aa8b732 | 1656 | goto out; |
3e4bb3ac XZ |
1657 | r = -EFAULT; |
1658 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1659 | goto out_free2; | |
1660 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1661 | if (r) |
3e4bb3ac | 1662 | goto out_free2; |
6aa8b732 | 1663 | r = 0; |
3e4bb3ac XZ |
1664 | out_free2: |
1665 | kfree(kvm_regs); | |
6aa8b732 AK |
1666 | break; |
1667 | } | |
1668 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1669 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1670 | r = -ENOMEM; | |
1671 | if (!kvm_sregs) | |
1672 | goto out; | |
1673 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1674 | if (r) |
1675 | goto out; | |
1676 | r = -EFAULT; | |
fa3795a7 | 1677 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1678 | goto out; |
1679 | r = 0; | |
1680 | break; | |
1681 | } | |
1682 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1683 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1684 | r = -ENOMEM; | |
1685 | if (!kvm_sregs) | |
1686 | goto out; | |
6aa8b732 | 1687 | r = -EFAULT; |
fa3795a7 | 1688 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1689 | goto out; |
fa3795a7 | 1690 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1691 | if (r) |
1692 | goto out; | |
1693 | r = 0; | |
1694 | break; | |
1695 | } | |
62d9f0db MT |
1696 | case KVM_GET_MP_STATE: { |
1697 | struct kvm_mp_state mp_state; | |
1698 | ||
1699 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1700 | if (r) | |
1701 | goto out; | |
1702 | r = -EFAULT; | |
1703 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1704 | goto out; | |
1705 | r = 0; | |
1706 | break; | |
1707 | } | |
1708 | case KVM_SET_MP_STATE: { | |
1709 | struct kvm_mp_state mp_state; | |
1710 | ||
1711 | r = -EFAULT; | |
1712 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1713 | goto out; | |
1714 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1715 | if (r) | |
1716 | goto out; | |
1717 | r = 0; | |
1718 | break; | |
1719 | } | |
6aa8b732 AK |
1720 | case KVM_TRANSLATE: { |
1721 | struct kvm_translation tr; | |
1722 | ||
1723 | r = -EFAULT; | |
2f366987 | 1724 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1725 | goto out; |
8b006791 | 1726 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1727 | if (r) |
1728 | goto out; | |
1729 | r = -EFAULT; | |
2f366987 | 1730 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1731 | goto out; |
1732 | r = 0; | |
1733 | break; | |
1734 | } | |
d0bfb940 JK |
1735 | case KVM_SET_GUEST_DEBUG: { |
1736 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1737 | |
1738 | r = -EFAULT; | |
2f366987 | 1739 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1740 | goto out; |
d0bfb940 | 1741 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1742 | if (r) |
1743 | goto out; | |
1744 | r = 0; | |
1745 | break; | |
1746 | } | |
1961d276 AK |
1747 | case KVM_SET_SIGNAL_MASK: { |
1748 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1749 | struct kvm_signal_mask kvm_sigmask; | |
1750 | sigset_t sigset, *p; | |
1751 | ||
1752 | p = NULL; | |
1753 | if (argp) { | |
1754 | r = -EFAULT; | |
1755 | if (copy_from_user(&kvm_sigmask, argp, | |
1756 | sizeof kvm_sigmask)) | |
1757 | goto out; | |
1758 | r = -EINVAL; | |
1759 | if (kvm_sigmask.len != sizeof sigset) | |
1760 | goto out; | |
1761 | r = -EFAULT; | |
1762 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1763 | sizeof sigset)) | |
1764 | goto out; | |
1765 | p = &sigset; | |
1766 | } | |
376d41ff | 1767 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1768 | break; |
1769 | } | |
b8836737 | 1770 | case KVM_GET_FPU: { |
fa3795a7 DH |
1771 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1772 | r = -ENOMEM; | |
1773 | if (!fpu) | |
1774 | goto out; | |
1775 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1776 | if (r) |
1777 | goto out; | |
1778 | r = -EFAULT; | |
fa3795a7 | 1779 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1780 | goto out; |
1781 | r = 0; | |
1782 | break; | |
1783 | } | |
1784 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1785 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1786 | r = -ENOMEM; | |
1787 | if (!fpu) | |
1788 | goto out; | |
b8836737 | 1789 | r = -EFAULT; |
fa3795a7 | 1790 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1791 | goto out; |
fa3795a7 | 1792 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1793 | if (r) |
1794 | goto out; | |
1795 | r = 0; | |
1796 | break; | |
1797 | } | |
bccf2150 | 1798 | default: |
313a3dc7 | 1799 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1800 | } |
1801 | out: | |
2122ff5e | 1802 | vcpu_put(vcpu); |
fa3795a7 DH |
1803 | kfree(fpu); |
1804 | kfree(kvm_sregs); | |
bccf2150 AK |
1805 | return r; |
1806 | } | |
1807 | ||
1808 | static long kvm_vm_ioctl(struct file *filp, | |
1809 | unsigned int ioctl, unsigned long arg) | |
1810 | { | |
1811 | struct kvm *kvm = filp->private_data; | |
1812 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1813 | int r; |
bccf2150 | 1814 | |
6d4e4c4f AK |
1815 | if (kvm->mm != current->mm) |
1816 | return -EIO; | |
bccf2150 AK |
1817 | switch (ioctl) { |
1818 | case KVM_CREATE_VCPU: | |
1819 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1820 | if (r < 0) | |
1821 | goto out; | |
1822 | break; | |
6fc138d2 IE |
1823 | case KVM_SET_USER_MEMORY_REGION: { |
1824 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1825 | ||
1826 | r = -EFAULT; | |
1827 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1828 | sizeof kvm_userspace_mem)) | |
1829 | goto out; | |
1830 | ||
1831 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1832 | if (r) |
1833 | goto out; | |
1834 | break; | |
1835 | } | |
1836 | case KVM_GET_DIRTY_LOG: { | |
1837 | struct kvm_dirty_log log; | |
1838 | ||
1839 | r = -EFAULT; | |
2f366987 | 1840 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1841 | goto out; |
2c6f5df9 | 1842 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1843 | if (r) |
1844 | goto out; | |
1845 | break; | |
1846 | } | |
5f94c174 LV |
1847 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1848 | case KVM_REGISTER_COALESCED_MMIO: { | |
1849 | struct kvm_coalesced_mmio_zone zone; | |
1850 | r = -EFAULT; | |
1851 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1852 | goto out; | |
5f94c174 LV |
1853 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
1854 | if (r) | |
1855 | goto out; | |
1856 | r = 0; | |
1857 | break; | |
1858 | } | |
1859 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1860 | struct kvm_coalesced_mmio_zone zone; | |
1861 | r = -EFAULT; | |
1862 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1863 | goto out; | |
5f94c174 LV |
1864 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
1865 | if (r) | |
1866 | goto out; | |
1867 | r = 0; | |
1868 | break; | |
1869 | } | |
1870 | #endif | |
721eecbf GH |
1871 | case KVM_IRQFD: { |
1872 | struct kvm_irqfd data; | |
1873 | ||
1874 | r = -EFAULT; | |
1875 | if (copy_from_user(&data, argp, sizeof data)) | |
1876 | goto out; | |
1877 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1878 | break; | |
1879 | } | |
d34e6b17 GH |
1880 | case KVM_IOEVENTFD: { |
1881 | struct kvm_ioeventfd data; | |
1882 | ||
1883 | r = -EFAULT; | |
1884 | if (copy_from_user(&data, argp, sizeof data)) | |
1885 | goto out; | |
1886 | r = kvm_ioeventfd(kvm, &data); | |
1887 | break; | |
1888 | } | |
73880c80 GN |
1889 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1890 | case KVM_SET_BOOT_CPU_ID: | |
1891 | r = 0; | |
894a9c55 | 1892 | mutex_lock(&kvm->lock); |
73880c80 GN |
1893 | if (atomic_read(&kvm->online_vcpus) != 0) |
1894 | r = -EBUSY; | |
1895 | else | |
1896 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1897 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1898 | break; |
1899 | #endif | |
f17abe9a | 1900 | default: |
1fe779f8 | 1901 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1902 | if (r == -ENOTTY) |
1903 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1904 | } |
1905 | out: | |
1906 | return r; | |
1907 | } | |
1908 | ||
6ff5894c AB |
1909 | #ifdef CONFIG_COMPAT |
1910 | struct compat_kvm_dirty_log { | |
1911 | __u32 slot; | |
1912 | __u32 padding1; | |
1913 | union { | |
1914 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1915 | __u64 padding2; | |
1916 | }; | |
1917 | }; | |
1918 | ||
1919 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1920 | unsigned int ioctl, unsigned long arg) | |
1921 | { | |
1922 | struct kvm *kvm = filp->private_data; | |
1923 | int r; | |
1924 | ||
1925 | if (kvm->mm != current->mm) | |
1926 | return -EIO; | |
1927 | switch (ioctl) { | |
1928 | case KVM_GET_DIRTY_LOG: { | |
1929 | struct compat_kvm_dirty_log compat_log; | |
1930 | struct kvm_dirty_log log; | |
1931 | ||
1932 | r = -EFAULT; | |
1933 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1934 | sizeof(compat_log))) | |
1935 | goto out; | |
1936 | log.slot = compat_log.slot; | |
1937 | log.padding1 = compat_log.padding1; | |
1938 | log.padding2 = compat_log.padding2; | |
1939 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
1940 | ||
1941 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
1942 | if (r) | |
1943 | goto out; | |
1944 | break; | |
1945 | } | |
1946 | default: | |
1947 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
1948 | } | |
1949 | ||
1950 | out: | |
1951 | return r; | |
1952 | } | |
1953 | #endif | |
1954 | ||
e4a533a4 | 1955 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1956 | { |
777b3f49 MT |
1957 | struct page *page[1]; |
1958 | unsigned long addr; | |
1959 | int npages; | |
1960 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1961 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1962 | |
777b3f49 MT |
1963 | addr = gfn_to_hva(kvm, gfn); |
1964 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1965 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1966 | |
1967 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1968 | NULL); | |
1969 | if (unlikely(npages != 1)) | |
e4a533a4 | 1970 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1971 | |
1972 | vmf->page = page[0]; | |
e4a533a4 | 1973 | return 0; |
f17abe9a AK |
1974 | } |
1975 | ||
f0f37e2f | 1976 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 1977 | .fault = kvm_vm_fault, |
f17abe9a AK |
1978 | }; |
1979 | ||
1980 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1981 | { | |
1982 | vma->vm_ops = &kvm_vm_vm_ops; | |
1983 | return 0; | |
1984 | } | |
1985 | ||
3d3aab1b | 1986 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1987 | .release = kvm_vm_release, |
1988 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
1989 | #ifdef CONFIG_COMPAT |
1990 | .compat_ioctl = kvm_vm_compat_ioctl, | |
1991 | #endif | |
f17abe9a | 1992 | .mmap = kvm_vm_mmap, |
6038f373 | 1993 | .llseek = noop_llseek, |
f17abe9a AK |
1994 | }; |
1995 | ||
1996 | static int kvm_dev_ioctl_create_vm(void) | |
1997 | { | |
aac87636 | 1998 | int r; |
f17abe9a AK |
1999 | struct kvm *kvm; |
2000 | ||
f17abe9a | 2001 | kvm = kvm_create_vm(); |
d6d28168 AK |
2002 | if (IS_ERR(kvm)) |
2003 | return PTR_ERR(kvm); | |
6ce5a090 TY |
2004 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2005 | r = kvm_coalesced_mmio_init(kvm); | |
2006 | if (r < 0) { | |
2007 | kvm_put_kvm(kvm); | |
2008 | return r; | |
2009 | } | |
2010 | #endif | |
aac87636 HC |
2011 | r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2012 | if (r < 0) | |
66c0b394 | 2013 | kvm_put_kvm(kvm); |
f17abe9a | 2014 | |
aac87636 | 2015 | return r; |
f17abe9a AK |
2016 | } |
2017 | ||
1a811b61 AK |
2018 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2019 | { | |
2020 | switch (arg) { | |
ca9edaee | 2021 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2022 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2023 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
2024 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2025 | case KVM_CAP_SET_BOOT_CPU_ID: | |
2026 | #endif | |
a9c7399d | 2027 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 2028 | return 1; |
399ec807 AK |
2029 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
2030 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 2031 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2032 | #endif |
1a811b61 AK |
2033 | default: |
2034 | break; | |
2035 | } | |
2036 | return kvm_dev_ioctl_check_extension(arg); | |
2037 | } | |
2038 | ||
f17abe9a AK |
2039 | static long kvm_dev_ioctl(struct file *filp, |
2040 | unsigned int ioctl, unsigned long arg) | |
2041 | { | |
07c45a36 | 2042 | long r = -EINVAL; |
f17abe9a AK |
2043 | |
2044 | switch (ioctl) { | |
2045 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2046 | r = -EINVAL; |
2047 | if (arg) | |
2048 | goto out; | |
f17abe9a AK |
2049 | r = KVM_API_VERSION; |
2050 | break; | |
2051 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2052 | r = -EINVAL; |
2053 | if (arg) | |
2054 | goto out; | |
f17abe9a AK |
2055 | r = kvm_dev_ioctl_create_vm(); |
2056 | break; | |
018d00d2 | 2057 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2058 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2059 | break; |
07c45a36 AK |
2060 | case KVM_GET_VCPU_MMAP_SIZE: |
2061 | r = -EINVAL; | |
2062 | if (arg) | |
2063 | goto out; | |
adb1ff46 AK |
2064 | r = PAGE_SIZE; /* struct kvm_run */ |
2065 | #ifdef CONFIG_X86 | |
2066 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2067 | #endif |
2068 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2069 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2070 | #endif |
07c45a36 | 2071 | break; |
d4c9ff2d FEL |
2072 | case KVM_TRACE_ENABLE: |
2073 | case KVM_TRACE_PAUSE: | |
2074 | case KVM_TRACE_DISABLE: | |
2023a29c | 2075 | r = -EOPNOTSUPP; |
d4c9ff2d | 2076 | break; |
6aa8b732 | 2077 | default: |
043405e1 | 2078 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2079 | } |
2080 | out: | |
2081 | return r; | |
2082 | } | |
2083 | ||
6aa8b732 | 2084 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2085 | .unlocked_ioctl = kvm_dev_ioctl, |
2086 | .compat_ioctl = kvm_dev_ioctl, | |
6038f373 | 2087 | .llseek = noop_llseek, |
6aa8b732 AK |
2088 | }; |
2089 | ||
2090 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2091 | KVM_MINOR, |
6aa8b732 AK |
2092 | "kvm", |
2093 | &kvm_chardev_ops, | |
2094 | }; | |
2095 | ||
75b7127c | 2096 | static void hardware_enable_nolock(void *junk) |
1b6c0168 AK |
2097 | { |
2098 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 2099 | int r; |
1b6c0168 | 2100 | |
7f59f492 | 2101 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2102 | return; |
10474ae8 | 2103 | |
7f59f492 | 2104 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
2105 | |
2106 | r = kvm_arch_hardware_enable(NULL); | |
2107 | ||
2108 | if (r) { | |
2109 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
2110 | atomic_inc(&hardware_enable_failed); | |
2111 | printk(KERN_INFO "kvm: enabling virtualization on " | |
2112 | "CPU%d failed\n", cpu); | |
2113 | } | |
1b6c0168 AK |
2114 | } |
2115 | ||
75b7127c TY |
2116 | static void hardware_enable(void *junk) |
2117 | { | |
2118 | spin_lock(&kvm_lock); | |
2119 | hardware_enable_nolock(junk); | |
2120 | spin_unlock(&kvm_lock); | |
2121 | } | |
2122 | ||
2123 | static void hardware_disable_nolock(void *junk) | |
1b6c0168 AK |
2124 | { |
2125 | int cpu = raw_smp_processor_id(); | |
2126 | ||
7f59f492 | 2127 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2128 | return; |
7f59f492 | 2129 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2130 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2131 | } |
2132 | ||
75b7127c TY |
2133 | static void hardware_disable(void *junk) |
2134 | { | |
2135 | spin_lock(&kvm_lock); | |
2136 | hardware_disable_nolock(junk); | |
2137 | spin_unlock(&kvm_lock); | |
2138 | } | |
2139 | ||
10474ae8 AG |
2140 | static void hardware_disable_all_nolock(void) |
2141 | { | |
2142 | BUG_ON(!kvm_usage_count); | |
2143 | ||
2144 | kvm_usage_count--; | |
2145 | if (!kvm_usage_count) | |
75b7127c | 2146 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
10474ae8 AG |
2147 | } |
2148 | ||
2149 | static void hardware_disable_all(void) | |
2150 | { | |
2151 | spin_lock(&kvm_lock); | |
2152 | hardware_disable_all_nolock(); | |
2153 | spin_unlock(&kvm_lock); | |
2154 | } | |
2155 | ||
2156 | static int hardware_enable_all(void) | |
2157 | { | |
2158 | int r = 0; | |
2159 | ||
2160 | spin_lock(&kvm_lock); | |
2161 | ||
2162 | kvm_usage_count++; | |
2163 | if (kvm_usage_count == 1) { | |
2164 | atomic_set(&hardware_enable_failed, 0); | |
75b7127c | 2165 | on_each_cpu(hardware_enable_nolock, NULL, 1); |
10474ae8 AG |
2166 | |
2167 | if (atomic_read(&hardware_enable_failed)) { | |
2168 | hardware_disable_all_nolock(); | |
2169 | r = -EBUSY; | |
2170 | } | |
2171 | } | |
2172 | ||
2173 | spin_unlock(&kvm_lock); | |
2174 | ||
2175 | return r; | |
2176 | } | |
2177 | ||
774c47f1 AK |
2178 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2179 | void *v) | |
2180 | { | |
2181 | int cpu = (long)v; | |
2182 | ||
10474ae8 AG |
2183 | if (!kvm_usage_count) |
2184 | return NOTIFY_OK; | |
2185 | ||
1a6f4d7f | 2186 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2187 | switch (val) { |
cec9ad27 | 2188 | case CPU_DYING: |
6ec8a856 AK |
2189 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2190 | cpu); | |
2191 | hardware_disable(NULL); | |
2192 | break; | |
da908f2f | 2193 | case CPU_STARTING: |
43934a38 JK |
2194 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2195 | cpu); | |
da908f2f | 2196 | hardware_enable(NULL); |
774c47f1 AK |
2197 | break; |
2198 | } | |
2199 | return NOTIFY_OK; | |
2200 | } | |
2201 | ||
4ecac3fd | 2202 | |
b7c4145b | 2203 | asmlinkage void kvm_spurious_fault(void) |
4ecac3fd | 2204 | { |
4ecac3fd AK |
2205 | /* Fault while not rebooting. We want the trace. */ |
2206 | BUG(); | |
2207 | } | |
b7c4145b | 2208 | EXPORT_SYMBOL_GPL(kvm_spurious_fault); |
4ecac3fd | 2209 | |
9a2b85c6 | 2210 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2211 | void *v) |
9a2b85c6 | 2212 | { |
8e1c1815 SY |
2213 | /* |
2214 | * Some (well, at least mine) BIOSes hang on reboot if | |
2215 | * in vmx root mode. | |
2216 | * | |
2217 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2218 | */ | |
2219 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2220 | kvm_rebooting = true; | |
75b7127c | 2221 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
9a2b85c6 RR |
2222 | return NOTIFY_OK; |
2223 | } | |
2224 | ||
2225 | static struct notifier_block kvm_reboot_notifier = { | |
2226 | .notifier_call = kvm_reboot, | |
2227 | .priority = 0, | |
2228 | }; | |
2229 | ||
e93f8a0f | 2230 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
2231 | { |
2232 | int i; | |
2233 | ||
2234 | for (i = 0; i < bus->dev_count; i++) { | |
2235 | struct kvm_io_device *pos = bus->devs[i]; | |
2236 | ||
2237 | kvm_iodevice_destructor(pos); | |
2238 | } | |
e93f8a0f | 2239 | kfree(bus); |
2eeb2e94 GH |
2240 | } |
2241 | ||
bda9020e | 2242 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2243 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2244 | int len, const void *val) |
2eeb2e94 GH |
2245 | { |
2246 | int i; | |
90d83dc3 LJ |
2247 | struct kvm_io_bus *bus; |
2248 | ||
2249 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
bda9020e MT |
2250 | for (i = 0; i < bus->dev_count; i++) |
2251 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
2252 | return 0; | |
2253 | return -EOPNOTSUPP; | |
2254 | } | |
2eeb2e94 | 2255 | |
bda9020e | 2256 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2257 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2258 | int len, void *val) | |
bda9020e MT |
2259 | { |
2260 | int i; | |
90d83dc3 | 2261 | struct kvm_io_bus *bus; |
e93f8a0f | 2262 | |
90d83dc3 | 2263 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
bda9020e MT |
2264 | for (i = 0; i < bus->dev_count; i++) |
2265 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
2266 | return 0; | |
2267 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
2268 | } |
2269 | ||
79fac95e | 2270 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2271 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2272 | struct kvm_io_device *dev) | |
6c474694 | 2273 | { |
e93f8a0f | 2274 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2275 | |
e93f8a0f | 2276 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2277 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2278 | return -ENOSPC; | |
2eeb2e94 | 2279 | |
e93f8a0f MT |
2280 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2281 | if (!new_bus) | |
2282 | return -ENOMEM; | |
2283 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2284 | new_bus->devs[new_bus->dev_count++] = dev; | |
2285 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2286 | synchronize_srcu_expedited(&kvm->srcu); | |
2287 | kfree(bus); | |
090b7aff GH |
2288 | |
2289 | return 0; | |
2290 | } | |
2291 | ||
79fac95e | 2292 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2293 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2294 | struct kvm_io_device *dev) | |
090b7aff | 2295 | { |
e93f8a0f MT |
2296 | int i, r; |
2297 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2298 | |
e93f8a0f MT |
2299 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2300 | if (!new_bus) | |
2301 | return -ENOMEM; | |
090b7aff | 2302 | |
e93f8a0f MT |
2303 | bus = kvm->buses[bus_idx]; |
2304 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2305 | ||
2306 | r = -ENOENT; | |
2307 | for (i = 0; i < new_bus->dev_count; i++) | |
2308 | if (new_bus->devs[i] == dev) { | |
2309 | r = 0; | |
2310 | new_bus->devs[i] = new_bus->devs[--new_bus->dev_count]; | |
090b7aff GH |
2311 | break; |
2312 | } | |
e93f8a0f MT |
2313 | |
2314 | if (r) { | |
2315 | kfree(new_bus); | |
2316 | return r; | |
2317 | } | |
2318 | ||
2319 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2320 | synchronize_srcu_expedited(&kvm->srcu); | |
2321 | kfree(bus); | |
2322 | return r; | |
2eeb2e94 GH |
2323 | } |
2324 | ||
774c47f1 AK |
2325 | static struct notifier_block kvm_cpu_notifier = { |
2326 | .notifier_call = kvm_cpu_hotplug, | |
774c47f1 AK |
2327 | }; |
2328 | ||
8b88b099 | 2329 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2330 | { |
2331 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2332 | struct kvm *kvm; |
2333 | ||
8b88b099 | 2334 | *val = 0; |
ba1389b7 AK |
2335 | spin_lock(&kvm_lock); |
2336 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2337 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2338 | spin_unlock(&kvm_lock); |
8b88b099 | 2339 | return 0; |
ba1389b7 AK |
2340 | } |
2341 | ||
2342 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2343 | ||
8b88b099 | 2344 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2345 | { |
2346 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2347 | struct kvm *kvm; |
2348 | struct kvm_vcpu *vcpu; | |
2349 | int i; | |
2350 | ||
8b88b099 | 2351 | *val = 0; |
1165f5fe AK |
2352 | spin_lock(&kvm_lock); |
2353 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
2354 | kvm_for_each_vcpu(i, vcpu, kvm) |
2355 | *val += *(u32 *)((void *)vcpu + offset); | |
2356 | ||
1165f5fe | 2357 | spin_unlock(&kvm_lock); |
8b88b099 | 2358 | return 0; |
1165f5fe AK |
2359 | } |
2360 | ||
ba1389b7 AK |
2361 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2362 | ||
828c0950 | 2363 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2364 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2365 | [KVM_STAT_VM] = &vm_stat_fops, | |
2366 | }; | |
1165f5fe | 2367 | |
a16b043c | 2368 | static void kvm_init_debug(void) |
6aa8b732 AK |
2369 | { |
2370 | struct kvm_stats_debugfs_item *p; | |
2371 | ||
76f7c879 | 2372 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2373 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2374 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2375 | (void *)(long)p->offset, |
ba1389b7 | 2376 | stat_fops[p->kind]); |
6aa8b732 AK |
2377 | } |
2378 | ||
2379 | static void kvm_exit_debug(void) | |
2380 | { | |
2381 | struct kvm_stats_debugfs_item *p; | |
2382 | ||
2383 | for (p = debugfs_entries; p->name; ++p) | |
2384 | debugfs_remove(p->dentry); | |
76f7c879 | 2385 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2386 | } |
2387 | ||
59ae6c6b AK |
2388 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2389 | { | |
10474ae8 | 2390 | if (kvm_usage_count) |
75b7127c | 2391 | hardware_disable_nolock(NULL); |
59ae6c6b AK |
2392 | return 0; |
2393 | } | |
2394 | ||
2395 | static int kvm_resume(struct sys_device *dev) | |
2396 | { | |
ca84d1a2 ZA |
2397 | if (kvm_usage_count) { |
2398 | WARN_ON(spin_is_locked(&kvm_lock)); | |
75b7127c | 2399 | hardware_enable_nolock(NULL); |
ca84d1a2 | 2400 | } |
59ae6c6b AK |
2401 | return 0; |
2402 | } | |
2403 | ||
2404 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2405 | .name = "kvm", |
59ae6c6b AK |
2406 | .suspend = kvm_suspend, |
2407 | .resume = kvm_resume, | |
2408 | }; | |
2409 | ||
2410 | static struct sys_device kvm_sysdev = { | |
2411 | .id = 0, | |
2412 | .cls = &kvm_sysdev_class, | |
2413 | }; | |
2414 | ||
cea7bb21 | 2415 | struct page *bad_page; |
35149e21 | 2416 | pfn_t bad_pfn; |
6aa8b732 | 2417 | |
15ad7146 AK |
2418 | static inline |
2419 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2420 | { | |
2421 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2422 | } | |
2423 | ||
2424 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2425 | { | |
2426 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2427 | ||
e9b11c17 | 2428 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2429 | } |
2430 | ||
2431 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2432 | struct task_struct *next) | |
2433 | { | |
2434 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2435 | ||
e9b11c17 | 2436 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2437 | } |
2438 | ||
0ee75bea | 2439 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2440 | struct module *module) |
6aa8b732 AK |
2441 | { |
2442 | int r; | |
002c7f7c | 2443 | int cpu; |
6aa8b732 | 2444 | |
f8c16bba ZX |
2445 | r = kvm_arch_init(opaque); |
2446 | if (r) | |
d2308784 | 2447 | goto out_fail; |
cb498ea2 ZX |
2448 | |
2449 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2450 | ||
2451 | if (bad_page == NULL) { | |
2452 | r = -ENOMEM; | |
2453 | goto out; | |
2454 | } | |
2455 | ||
35149e21 AL |
2456 | bad_pfn = page_to_pfn(bad_page); |
2457 | ||
bf998156 HY |
2458 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2459 | ||
2460 | if (hwpoison_page == NULL) { | |
2461 | r = -ENOMEM; | |
2462 | goto out_free_0; | |
2463 | } | |
2464 | ||
2465 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2466 | ||
edba23e5 GN |
2467 | fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2468 | ||
2469 | if (fault_page == NULL) { | |
2470 | r = -ENOMEM; | |
2471 | goto out_free_0; | |
2472 | } | |
2473 | ||
2474 | fault_pfn = page_to_pfn(fault_page); | |
2475 | ||
8437a617 | 2476 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2477 | r = -ENOMEM; |
2478 | goto out_free_0; | |
2479 | } | |
2480 | ||
e9b11c17 | 2481 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2482 | if (r < 0) |
7f59f492 | 2483 | goto out_free_0a; |
6aa8b732 | 2484 | |
002c7f7c YS |
2485 | for_each_online_cpu(cpu) { |
2486 | smp_call_function_single(cpu, | |
e9b11c17 | 2487 | kvm_arch_check_processor_compat, |
8691e5a8 | 2488 | &r, 1); |
002c7f7c | 2489 | if (r < 0) |
d2308784 | 2490 | goto out_free_1; |
002c7f7c YS |
2491 | } |
2492 | ||
774c47f1 AK |
2493 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2494 | if (r) | |
d2308784 | 2495 | goto out_free_2; |
6aa8b732 AK |
2496 | register_reboot_notifier(&kvm_reboot_notifier); |
2497 | ||
59ae6c6b AK |
2498 | r = sysdev_class_register(&kvm_sysdev_class); |
2499 | if (r) | |
d2308784 | 2500 | goto out_free_3; |
59ae6c6b AK |
2501 | |
2502 | r = sysdev_register(&kvm_sysdev); | |
2503 | if (r) | |
d2308784 | 2504 | goto out_free_4; |
59ae6c6b | 2505 | |
c16f862d | 2506 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2507 | if (!vcpu_align) |
2508 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2509 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2510 | 0, NULL); |
c16f862d RR |
2511 | if (!kvm_vcpu_cache) { |
2512 | r = -ENOMEM; | |
d2308784 | 2513 | goto out_free_5; |
c16f862d RR |
2514 | } |
2515 | ||
af585b92 GN |
2516 | r = kvm_async_pf_init(); |
2517 | if (r) | |
2518 | goto out_free; | |
2519 | ||
6aa8b732 | 2520 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2521 | kvm_vm_fops.owner = module; |
2522 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2523 | |
2524 | r = misc_register(&kvm_dev); | |
2525 | if (r) { | |
d77c26fc | 2526 | printk(KERN_ERR "kvm: misc device register failed\n"); |
af585b92 | 2527 | goto out_unreg; |
6aa8b732 AK |
2528 | } |
2529 | ||
15ad7146 AK |
2530 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2531 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2532 | ||
0ea4ed8e DW |
2533 | kvm_init_debug(); |
2534 | ||
c7addb90 | 2535 | return 0; |
6aa8b732 | 2536 | |
af585b92 GN |
2537 | out_unreg: |
2538 | kvm_async_pf_deinit(); | |
6aa8b732 | 2539 | out_free: |
c16f862d | 2540 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2541 | out_free_5: |
59ae6c6b | 2542 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2543 | out_free_4: |
59ae6c6b | 2544 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2545 | out_free_3: |
6aa8b732 | 2546 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2547 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2548 | out_free_2: |
d2308784 | 2549 | out_free_1: |
e9b11c17 | 2550 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2551 | out_free_0a: |
2552 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2553 | out_free_0: |
edba23e5 GN |
2554 | if (fault_page) |
2555 | __free_page(fault_page); | |
bf998156 HY |
2556 | if (hwpoison_page) |
2557 | __free_page(hwpoison_page); | |
d2308784 | 2558 | __free_page(bad_page); |
ca45aaae | 2559 | out: |
f8c16bba | 2560 | kvm_arch_exit(); |
d2308784 | 2561 | out_fail: |
6aa8b732 AK |
2562 | return r; |
2563 | } | |
cb498ea2 | 2564 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2565 | |
cb498ea2 | 2566 | void kvm_exit(void) |
6aa8b732 | 2567 | { |
0ea4ed8e | 2568 | kvm_exit_debug(); |
6aa8b732 | 2569 | misc_deregister(&kvm_dev); |
c16f862d | 2570 | kmem_cache_destroy(kvm_vcpu_cache); |
af585b92 | 2571 | kvm_async_pf_deinit(); |
59ae6c6b AK |
2572 | sysdev_unregister(&kvm_sysdev); |
2573 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2574 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2575 | unregister_cpu_notifier(&kvm_cpu_notifier); |
75b7127c | 2576 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
e9b11c17 | 2577 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2578 | kvm_arch_exit(); |
7f59f492 | 2579 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2580 | __free_page(hwpoison_page); |
cea7bb21 | 2581 | __free_page(bad_page); |
6aa8b732 | 2582 | } |
cb498ea2 | 2583 | EXPORT_SYMBOL_GPL(kvm_exit); |