Commit | Line | Data |
---|---|---|
6aa8b732 AK |
1 | /* |
2 | * Kernel-based Virtual Machine driver for Linux | |
3 | * | |
4 | * This module enables machines with Intel VT-x extensions to run virtual | |
5 | * machines without emulation or binary translation. | |
6 | * | |
7 | * Copyright (C) 2006 Qumranet, Inc. | |
8 | * | |
9 | * Authors: | |
10 | * Avi Kivity <avi@qumranet.com> | |
11 | * Yaniv Kamay <yaniv@qumranet.com> | |
12 | * | |
13 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
14 | * the COPYING file in the top-level directory. | |
15 | * | |
16 | */ | |
17 | ||
e2174021 | 18 | #include "iodev.h" |
6aa8b732 | 19 | |
edf88417 | 20 | #include <linux/kvm_host.h> |
6aa8b732 AK |
21 | #include <linux/kvm.h> |
22 | #include <linux/module.h> | |
23 | #include <linux/errno.h> | |
6aa8b732 | 24 | #include <linux/percpu.h> |
6aa8b732 AK |
25 | #include <linux/mm.h> |
26 | #include <linux/miscdevice.h> | |
27 | #include <linux/vmalloc.h> | |
6aa8b732 | 28 | #include <linux/reboot.h> |
6aa8b732 AK |
29 | #include <linux/debugfs.h> |
30 | #include <linux/highmem.h> | |
31 | #include <linux/file.h> | |
59ae6c6b | 32 | #include <linux/sysdev.h> |
774c47f1 | 33 | #include <linux/cpu.h> |
e8edc6e0 | 34 | #include <linux/sched.h> |
d9e368d6 AK |
35 | #include <linux/cpumask.h> |
36 | #include <linux/smp.h> | |
d6d28168 | 37 | #include <linux/anon_inodes.h> |
04d2cc77 | 38 | #include <linux/profile.h> |
7aa81cc0 | 39 | #include <linux/kvm_para.h> |
6fc138d2 | 40 | #include <linux/pagemap.h> |
8d4e1288 | 41 | #include <linux/mman.h> |
35149e21 | 42 | #include <linux/swap.h> |
e56d532f | 43 | #include <linux/bitops.h> |
547de29e | 44 | #include <linux/spinlock.h> |
6ff5894c | 45 | #include <linux/compat.h> |
bc6678a3 | 46 | #include <linux/srcu.h> |
8f0b1ab6 | 47 | #include <linux/hugetlb.h> |
5a0e3ad6 | 48 | #include <linux/slab.h> |
6aa8b732 | 49 | |
e495606d | 50 | #include <asm/processor.h> |
e495606d AK |
51 | #include <asm/io.h> |
52 | #include <asm/uaccess.h> | |
3e021bf5 | 53 | #include <asm/pgtable.h> |
c8240bd6 | 54 | #include <asm-generic/bitops/le.h> |
6aa8b732 | 55 | |
5f94c174 | 56 | #include "coalesced_mmio.h" |
5f94c174 | 57 | |
229456fc MT |
58 | #define CREATE_TRACE_POINTS |
59 | #include <trace/events/kvm.h> | |
60 | ||
6aa8b732 AK |
61 | MODULE_AUTHOR("Qumranet"); |
62 | MODULE_LICENSE("GPL"); | |
63 | ||
fa40a821 MT |
64 | /* |
65 | * Ordering of locks: | |
66 | * | |
fae3a353 | 67 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
68 | */ |
69 | ||
e9b11c17 ZX |
70 | DEFINE_SPINLOCK(kvm_lock); |
71 | LIST_HEAD(vm_list); | |
133de902 | 72 | |
7f59f492 | 73 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
74 | static int kvm_usage_count = 0; |
75 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 76 | |
c16f862d RR |
77 | struct kmem_cache *kvm_vcpu_cache; |
78 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 79 | |
15ad7146 AK |
80 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
81 | ||
76f7c879 | 82 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 83 | |
bccf2150 AK |
84 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
85 | unsigned long arg); | |
10474ae8 AG |
86 | static int hardware_enable_all(void); |
87 | static void hardware_disable_all(void); | |
bccf2150 | 88 | |
e93f8a0f MT |
89 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
90 | ||
e8ba5d31 | 91 | static bool kvm_rebooting; |
4ecac3fd | 92 | |
54dee993 MT |
93 | static bool largepages_enabled = true; |
94 | ||
bf998156 HY |
95 | struct page *hwpoison_page; |
96 | pfn_t hwpoison_pfn; | |
97 | ||
c77fb9dc | 98 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 99 | { |
fc5659c8 JR |
100 | if (pfn_valid(pfn)) { |
101 | struct page *page = compound_head(pfn_to_page(pfn)); | |
102 | return PageReserved(page); | |
103 | } | |
cbff90a7 BAY |
104 | |
105 | return true; | |
106 | } | |
107 | ||
bccf2150 AK |
108 | /* |
109 | * Switches to specified vcpu, until a matching vcpu_put() | |
110 | */ | |
313a3dc7 | 111 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 112 | { |
15ad7146 AK |
113 | int cpu; |
114 | ||
bccf2150 | 115 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
116 | cpu = get_cpu(); |
117 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 118 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 119 | put_cpu(); |
6aa8b732 AK |
120 | } |
121 | ||
313a3dc7 | 122 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 123 | { |
15ad7146 | 124 | preempt_disable(); |
313a3dc7 | 125 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
126 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
127 | preempt_enable(); | |
6aa8b732 AK |
128 | mutex_unlock(&vcpu->mutex); |
129 | } | |
130 | ||
d9e368d6 AK |
131 | static void ack_flush(void *_completed) |
132 | { | |
d9e368d6 AK |
133 | } |
134 | ||
49846896 | 135 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 136 | { |
597a5f55 | 137 | int i, cpu, me; |
6ef7a1bc RR |
138 | cpumask_var_t cpus; |
139 | bool called = true; | |
d9e368d6 | 140 | struct kvm_vcpu *vcpu; |
d9e368d6 | 141 | |
79f55997 | 142 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 143 | |
70e335e1 | 144 | raw_spin_lock(&kvm->requests_lock); |
e601e3be | 145 | me = smp_processor_id(); |
988a2cae | 146 | kvm_for_each_vcpu(i, vcpu, kvm) { |
49846896 | 147 | if (test_and_set_bit(req, &vcpu->requests)) |
d9e368d6 AK |
148 | continue; |
149 | cpu = vcpu->cpu; | |
6ef7a1bc RR |
150 | if (cpus != NULL && cpu != -1 && cpu != me) |
151 | cpumask_set_cpu(cpu, cpus); | |
49846896 | 152 | } |
6ef7a1bc RR |
153 | if (unlikely(cpus == NULL)) |
154 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
155 | else if (!cpumask_empty(cpus)) | |
156 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
157 | else | |
158 | called = false; | |
70e335e1 | 159 | raw_spin_unlock(&kvm->requests_lock); |
6ef7a1bc | 160 | free_cpumask_var(cpus); |
49846896 | 161 | return called; |
d9e368d6 AK |
162 | } |
163 | ||
49846896 | 164 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 165 | { |
49846896 RR |
166 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
167 | ++kvm->stat.remote_tlb_flush; | |
2e53d63a MT |
168 | } |
169 | ||
49846896 RR |
170 | void kvm_reload_remote_mmus(struct kvm *kvm) |
171 | { | |
172 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
173 | } | |
2e53d63a | 174 | |
fb3f0f51 RR |
175 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
176 | { | |
177 | struct page *page; | |
178 | int r; | |
179 | ||
180 | mutex_init(&vcpu->mutex); | |
181 | vcpu->cpu = -1; | |
fb3f0f51 RR |
182 | vcpu->kvm = kvm; |
183 | vcpu->vcpu_id = id; | |
b6958ce4 | 184 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
185 | |
186 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
187 | if (!page) { | |
188 | r = -ENOMEM; | |
189 | goto fail; | |
190 | } | |
191 | vcpu->run = page_address(page); | |
192 | ||
e9b11c17 | 193 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 194 | if (r < 0) |
e9b11c17 | 195 | goto fail_free_run; |
fb3f0f51 RR |
196 | return 0; |
197 | ||
fb3f0f51 RR |
198 | fail_free_run: |
199 | free_page((unsigned long)vcpu->run); | |
200 | fail: | |
76fafa5e | 201 | return r; |
fb3f0f51 RR |
202 | } |
203 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
204 | ||
205 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
206 | { | |
e9b11c17 | 207 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
208 | free_page((unsigned long)vcpu->run); |
209 | } | |
210 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
211 | ||
e930bffe AA |
212 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
213 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
214 | { | |
215 | return container_of(mn, struct kvm, mmu_notifier); | |
216 | } | |
217 | ||
218 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
219 | struct mm_struct *mm, | |
220 | unsigned long address) | |
221 | { | |
222 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 223 | int need_tlb_flush, idx; |
e930bffe AA |
224 | |
225 | /* | |
226 | * When ->invalidate_page runs, the linux pte has been zapped | |
227 | * already but the page is still allocated until | |
228 | * ->invalidate_page returns. So if we increase the sequence | |
229 | * here the kvm page fault will notice if the spte can't be | |
230 | * established because the page is going to be freed. If | |
231 | * instead the kvm page fault establishes the spte before | |
232 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
233 | * before returning. | |
234 | * | |
235 | * The sequence increase only need to be seen at spin_unlock | |
236 | * time, and not at spin_lock time. | |
237 | * | |
238 | * Increasing the sequence after the spin_unlock would be | |
239 | * unsafe because the kvm page fault could then establish the | |
240 | * pte after kvm_unmap_hva returned, without noticing the page | |
241 | * is going to be freed. | |
242 | */ | |
bc6678a3 | 243 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
244 | spin_lock(&kvm->mmu_lock); |
245 | kvm->mmu_notifier_seq++; | |
246 | need_tlb_flush = kvm_unmap_hva(kvm, address); | |
247 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 248 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
249 | |
250 | /* we've to flush the tlb before the pages can be freed */ | |
251 | if (need_tlb_flush) | |
252 | kvm_flush_remote_tlbs(kvm); | |
253 | ||
254 | } | |
255 | ||
3da0dd43 IE |
256 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
257 | struct mm_struct *mm, | |
258 | unsigned long address, | |
259 | pte_t pte) | |
260 | { | |
261 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 262 | int idx; |
3da0dd43 | 263 | |
bc6678a3 | 264 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
265 | spin_lock(&kvm->mmu_lock); |
266 | kvm->mmu_notifier_seq++; | |
267 | kvm_set_spte_hva(kvm, address, pte); | |
268 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 269 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
270 | } |
271 | ||
e930bffe AA |
272 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
273 | struct mm_struct *mm, | |
274 | unsigned long start, | |
275 | unsigned long end) | |
276 | { | |
277 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 278 | int need_tlb_flush = 0, idx; |
e930bffe | 279 | |
bc6678a3 | 280 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
281 | spin_lock(&kvm->mmu_lock); |
282 | /* | |
283 | * The count increase must become visible at unlock time as no | |
284 | * spte can be established without taking the mmu_lock and | |
285 | * count is also read inside the mmu_lock critical section. | |
286 | */ | |
287 | kvm->mmu_notifier_count++; | |
288 | for (; start < end; start += PAGE_SIZE) | |
289 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
290 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 291 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
292 | |
293 | /* we've to flush the tlb before the pages can be freed */ | |
294 | if (need_tlb_flush) | |
295 | kvm_flush_remote_tlbs(kvm); | |
296 | } | |
297 | ||
298 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
299 | struct mm_struct *mm, | |
300 | unsigned long start, | |
301 | unsigned long end) | |
302 | { | |
303 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
304 | ||
305 | spin_lock(&kvm->mmu_lock); | |
306 | /* | |
307 | * This sequence increase will notify the kvm page fault that | |
308 | * the page that is going to be mapped in the spte could have | |
309 | * been freed. | |
310 | */ | |
311 | kvm->mmu_notifier_seq++; | |
312 | /* | |
313 | * The above sequence increase must be visible before the | |
314 | * below count decrease but both values are read by the kvm | |
315 | * page fault under mmu_lock spinlock so we don't need to add | |
316 | * a smb_wmb() here in between the two. | |
317 | */ | |
318 | kvm->mmu_notifier_count--; | |
319 | spin_unlock(&kvm->mmu_lock); | |
320 | ||
321 | BUG_ON(kvm->mmu_notifier_count < 0); | |
322 | } | |
323 | ||
324 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
325 | struct mm_struct *mm, | |
326 | unsigned long address) | |
327 | { | |
328 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 329 | int young, idx; |
e930bffe | 330 | |
bc6678a3 | 331 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
332 | spin_lock(&kvm->mmu_lock); |
333 | young = kvm_age_hva(kvm, address); | |
334 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 335 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
336 | |
337 | if (young) | |
338 | kvm_flush_remote_tlbs(kvm); | |
339 | ||
340 | return young; | |
341 | } | |
342 | ||
85db06e5 MT |
343 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
344 | struct mm_struct *mm) | |
345 | { | |
346 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
347 | int idx; |
348 | ||
349 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 350 | kvm_arch_flush_shadow(kvm); |
eda2beda | 351 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
352 | } |
353 | ||
e930bffe AA |
354 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
355 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
356 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
357 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
358 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
3da0dd43 | 359 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 360 | .release = kvm_mmu_notifier_release, |
e930bffe | 361 | }; |
4c07b0a4 AK |
362 | |
363 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
364 | { | |
365 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
366 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
367 | } | |
368 | ||
369 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
370 | ||
371 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
372 | { | |
373 | return 0; | |
374 | } | |
375 | ||
e930bffe AA |
376 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
377 | ||
f17abe9a | 378 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 379 | { |
e93f8a0f | 380 | int r = 0, i; |
d19a9cd2 | 381 | struct kvm *kvm = kvm_arch_create_vm(); |
6aa8b732 | 382 | |
d19a9cd2 ZX |
383 | if (IS_ERR(kvm)) |
384 | goto out; | |
10474ae8 AG |
385 | |
386 | r = hardware_enable_all(); | |
387 | if (r) | |
388 | goto out_err_nodisable; | |
389 | ||
75858a84 AK |
390 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
391 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 392 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 393 | #endif |
6aa8b732 | 394 | |
46a26bf5 MT |
395 | r = -ENOMEM; |
396 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
397 | if (!kvm->memslots) | |
398 | goto out_err; | |
bc6678a3 MT |
399 | if (init_srcu_struct(&kvm->srcu)) |
400 | goto out_err; | |
e93f8a0f MT |
401 | for (i = 0; i < KVM_NR_BUSES; i++) { |
402 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
403 | GFP_KERNEL); | |
404 | if (!kvm->buses[i]) { | |
405 | cleanup_srcu_struct(&kvm->srcu); | |
406 | goto out_err; | |
407 | } | |
408 | } | |
46a26bf5 | 409 | |
4c07b0a4 | 410 | r = kvm_init_mmu_notifier(kvm); |
283d0c65 | 411 | if (r) { |
bc6678a3 | 412 | cleanup_srcu_struct(&kvm->srcu); |
283d0c65 | 413 | goto out_err; |
e930bffe | 414 | } |
e930bffe | 415 | |
6d4e4c4f AK |
416 | kvm->mm = current->mm; |
417 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 418 | spin_lock_init(&kvm->mmu_lock); |
70e335e1 | 419 | raw_spin_lock_init(&kvm->requests_lock); |
d34e6b17 | 420 | kvm_eventfd_init(kvm); |
11ec2804 | 421 | mutex_init(&kvm->lock); |
60eead79 | 422 | mutex_init(&kvm->irq_lock); |
79fac95e | 423 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 424 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
425 | spin_lock(&kvm_lock); |
426 | list_add(&kvm->vm_list, &vm_list); | |
427 | spin_unlock(&kvm_lock); | |
d19a9cd2 | 428 | out: |
f17abe9a | 429 | return kvm; |
10474ae8 AG |
430 | |
431 | out_err: | |
432 | hardware_disable_all(); | |
433 | out_err_nodisable: | |
e93f8a0f MT |
434 | for (i = 0; i < KVM_NR_BUSES; i++) |
435 | kfree(kvm->buses[i]); | |
46a26bf5 | 436 | kfree(kvm->memslots); |
10474ae8 AG |
437 | kfree(kvm); |
438 | return ERR_PTR(r); | |
f17abe9a AK |
439 | } |
440 | ||
6aa8b732 AK |
441 | /* |
442 | * Free any memory in @free but not in @dont. | |
443 | */ | |
444 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
445 | struct kvm_memory_slot *dont) | |
446 | { | |
ec04b260 JR |
447 | int i; |
448 | ||
290fc38d IE |
449 | if (!dont || free->rmap != dont->rmap) |
450 | vfree(free->rmap); | |
6aa8b732 AK |
451 | |
452 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
453 | vfree(free->dirty_bitmap); | |
454 | ||
ec04b260 JR |
455 | |
456 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
457 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
458 | vfree(free->lpage_info[i]); | |
459 | free->lpage_info[i] = NULL; | |
460 | } | |
461 | } | |
05da4558 | 462 | |
6aa8b732 | 463 | free->npages = 0; |
8b6d44c7 | 464 | free->dirty_bitmap = NULL; |
8d4e1288 | 465 | free->rmap = NULL; |
6aa8b732 AK |
466 | } |
467 | ||
d19a9cd2 | 468 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
469 | { |
470 | int i; | |
46a26bf5 MT |
471 | struct kvm_memslots *slots = kvm->memslots; |
472 | ||
473 | for (i = 0; i < slots->nmemslots; ++i) | |
474 | kvm_free_physmem_slot(&slots->memslots[i], NULL); | |
6aa8b732 | 475 | |
46a26bf5 | 476 | kfree(kvm->memslots); |
6aa8b732 AK |
477 | } |
478 | ||
f17abe9a AK |
479 | static void kvm_destroy_vm(struct kvm *kvm) |
480 | { | |
e93f8a0f | 481 | int i; |
6d4e4c4f AK |
482 | struct mm_struct *mm = kvm->mm; |
483 | ||
ad8ba2cd | 484 | kvm_arch_sync_events(kvm); |
133de902 AK |
485 | spin_lock(&kvm_lock); |
486 | list_del(&kvm->vm_list); | |
487 | spin_unlock(&kvm_lock); | |
399ec807 | 488 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
489 | for (i = 0; i < KVM_NR_BUSES; i++) |
490 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 491 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
492 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
493 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
494 | #else |
495 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 496 | #endif |
d19a9cd2 | 497 | kvm_arch_destroy_vm(kvm); |
10474ae8 | 498 | hardware_disable_all(); |
6d4e4c4f | 499 | mmdrop(mm); |
f17abe9a AK |
500 | } |
501 | ||
d39f13b0 IE |
502 | void kvm_get_kvm(struct kvm *kvm) |
503 | { | |
504 | atomic_inc(&kvm->users_count); | |
505 | } | |
506 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
507 | ||
508 | void kvm_put_kvm(struct kvm *kvm) | |
509 | { | |
510 | if (atomic_dec_and_test(&kvm->users_count)) | |
511 | kvm_destroy_vm(kvm); | |
512 | } | |
513 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
514 | ||
515 | ||
f17abe9a AK |
516 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
517 | { | |
518 | struct kvm *kvm = filp->private_data; | |
519 | ||
721eecbf GH |
520 | kvm_irqfd_release(kvm); |
521 | ||
d39f13b0 | 522 | kvm_put_kvm(kvm); |
6aa8b732 AK |
523 | return 0; |
524 | } | |
525 | ||
6aa8b732 AK |
526 | /* |
527 | * Allocate some memory and give it an address in the guest physical address | |
528 | * space. | |
529 | * | |
530 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 531 | * |
10589a46 | 532 | * Must be called holding mmap_sem for write. |
6aa8b732 | 533 | */ |
f78e0e2e SY |
534 | int __kvm_set_memory_region(struct kvm *kvm, |
535 | struct kvm_userspace_memory_region *mem, | |
536 | int user_alloc) | |
6aa8b732 | 537 | { |
bc6678a3 | 538 | int r, flush_shadow = 0; |
6aa8b732 | 539 | gfn_t base_gfn; |
28bcb112 HC |
540 | unsigned long npages; |
541 | unsigned long i; | |
6aa8b732 AK |
542 | struct kvm_memory_slot *memslot; |
543 | struct kvm_memory_slot old, new; | |
bc6678a3 | 544 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
545 | |
546 | r = -EINVAL; | |
547 | /* General sanity checks */ | |
548 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
549 | goto out; | |
550 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
551 | goto out; | |
e7cacd40 | 552 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 553 | goto out; |
e0d62c7f | 554 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
555 | goto out; |
556 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
557 | goto out; | |
558 | ||
46a26bf5 | 559 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
560 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
561 | npages = mem->memory_size >> PAGE_SHIFT; | |
562 | ||
660c22c4 TY |
563 | r = -EINVAL; |
564 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
565 | goto out; | |
566 | ||
6aa8b732 AK |
567 | if (!npages) |
568 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
569 | ||
6aa8b732 AK |
570 | new = old = *memslot; |
571 | ||
572 | new.base_gfn = base_gfn; | |
573 | new.npages = npages; | |
574 | new.flags = mem->flags; | |
575 | ||
576 | /* Disallow changing a memory slot's size. */ | |
577 | r = -EINVAL; | |
578 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 579 | goto out_free; |
6aa8b732 AK |
580 | |
581 | /* Check for overlaps */ | |
582 | r = -EEXIST; | |
583 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 584 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 585 | |
4cd481f6 | 586 | if (s == memslot || !s->npages) |
6aa8b732 AK |
587 | continue; |
588 | if (!((base_gfn + npages <= s->base_gfn) || | |
589 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 590 | goto out_free; |
6aa8b732 | 591 | } |
6aa8b732 | 592 | |
6aa8b732 AK |
593 | /* Free page dirty bitmap if unneeded */ |
594 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 595 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
596 | |
597 | r = -ENOMEM; | |
598 | ||
599 | /* Allocate if a slot is being created */ | |
eff0114a | 600 | #ifndef CONFIG_S390 |
8d4e1288 | 601 | if (npages && !new.rmap) { |
d77c26fc | 602 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
603 | |
604 | if (!new.rmap) | |
f78e0e2e | 605 | goto out_free; |
290fc38d | 606 | |
290fc38d | 607 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 608 | |
80b14b5b | 609 | new.user_alloc = user_alloc; |
bc6678a3 | 610 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 611 | } |
ec04b260 JR |
612 | if (!npages) |
613 | goto skip_lpage; | |
05da4558 | 614 | |
ec04b260 | 615 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
616 | unsigned long ugfn; |
617 | unsigned long j; | |
618 | int lpages; | |
ec04b260 | 619 | int level = i + 2; |
05da4558 | 620 | |
ec04b260 JR |
621 | /* Avoid unused variable warning if no large pages */ |
622 | (void)level; | |
623 | ||
624 | if (new.lpage_info[i]) | |
625 | continue; | |
626 | ||
627 | lpages = 1 + (base_gfn + npages - 1) / | |
628 | KVM_PAGES_PER_HPAGE(level); | |
629 | lpages -= base_gfn / KVM_PAGES_PER_HPAGE(level); | |
630 | ||
631 | new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i])); | |
632 | ||
633 | if (!new.lpage_info[i]) | |
05da4558 MT |
634 | goto out_free; |
635 | ||
ec04b260 JR |
636 | memset(new.lpage_info[i], 0, |
637 | lpages * sizeof(*new.lpage_info[i])); | |
05da4558 | 638 | |
ec04b260 JR |
639 | if (base_gfn % KVM_PAGES_PER_HPAGE(level)) |
640 | new.lpage_info[i][0].write_count = 1; | |
641 | if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE(level)) | |
642 | new.lpage_info[i][lpages - 1].write_count = 1; | |
ac04527f AK |
643 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
644 | /* | |
645 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
646 | * other, or if explicitly asked to, disable large page |
647 | * support for this slot | |
ac04527f | 648 | */ |
ec04b260 | 649 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 650 | !largepages_enabled) |
ec04b260 JR |
651 | for (j = 0; j < lpages; ++j) |
652 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 653 | } |
6aa8b732 | 654 | |
ec04b260 JR |
655 | skip_lpage: |
656 | ||
6aa8b732 AK |
657 | /* Allocate page dirty bitmap if needed */ |
658 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
87bf6e7d | 659 | unsigned long dirty_bytes = kvm_dirty_bitmap_bytes(&new); |
6aa8b732 AK |
660 | |
661 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
662 | if (!new.dirty_bitmap) | |
f78e0e2e | 663 | goto out_free; |
6aa8b732 | 664 | memset(new.dirty_bitmap, 0, dirty_bytes); |
bc6678a3 | 665 | /* destroy any largepage mappings for dirty tracking */ |
e244584f | 666 | if (old.npages) |
bc6678a3 | 667 | flush_shadow = 1; |
6aa8b732 | 668 | } |
3eea8437 CB |
669 | #else /* not defined CONFIG_S390 */ |
670 | new.user_alloc = user_alloc; | |
671 | if (user_alloc) | |
672 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 673 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 674 | |
bc6678a3 MT |
675 | if (!npages) { |
676 | r = -ENOMEM; | |
677 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
678 | if (!slots) | |
679 | goto out_free; | |
680 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
681 | if (mem->slot >= slots->nmemslots) | |
682 | slots->nmemslots = mem->slot + 1; | |
683 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; | |
684 | ||
685 | old_memslots = kvm->memslots; | |
686 | rcu_assign_pointer(kvm->memslots, slots); | |
687 | synchronize_srcu_expedited(&kvm->srcu); | |
688 | /* From this point no new shadow pages pointing to a deleted | |
689 | * memslot will be created. | |
690 | * | |
691 | * validation of sp->gfn happens in: | |
692 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
693 | * - kvm_is_visible_gfn (mmu_check_roots) | |
694 | */ | |
34d4cb8f | 695 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
696 | kfree(old_memslots); |
697 | } | |
34d4cb8f | 698 | |
f7784b8e MT |
699 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
700 | if (r) | |
701 | goto out_free; | |
702 | ||
bc6678a3 MT |
703 | #ifdef CONFIG_DMAR |
704 | /* map the pages in iommu page table */ | |
705 | if (npages) { | |
706 | r = kvm_iommu_map_pages(kvm, &new); | |
707 | if (r) | |
708 | goto out_free; | |
709 | } | |
710 | #endif | |
604b38ac | 711 | |
bc6678a3 MT |
712 | r = -ENOMEM; |
713 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
714 | if (!slots) | |
715 | goto out_free; | |
716 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
717 | if (mem->slot >= slots->nmemslots) | |
718 | slots->nmemslots = mem->slot + 1; | |
719 | ||
720 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
721 | if (!npages) { | |
722 | new.rmap = NULL; | |
723 | new.dirty_bitmap = NULL; | |
724 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
725 | new.lpage_info[i] = NULL; | |
726 | } | |
727 | ||
728 | slots->memslots[mem->slot] = new; | |
729 | old_memslots = kvm->memslots; | |
730 | rcu_assign_pointer(kvm->memslots, slots); | |
731 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 732 | |
f7784b8e | 733 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 734 | |
bc6678a3 MT |
735 | kvm_free_physmem_slot(&old, &new); |
736 | kfree(old_memslots); | |
737 | ||
738 | if (flush_shadow) | |
739 | kvm_arch_flush_shadow(kvm); | |
740 | ||
6aa8b732 AK |
741 | return 0; |
742 | ||
f78e0e2e | 743 | out_free: |
6aa8b732 AK |
744 | kvm_free_physmem_slot(&new, &old); |
745 | out: | |
746 | return r; | |
210c7c4d IE |
747 | |
748 | } | |
f78e0e2e SY |
749 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
750 | ||
751 | int kvm_set_memory_region(struct kvm *kvm, | |
752 | struct kvm_userspace_memory_region *mem, | |
753 | int user_alloc) | |
754 | { | |
755 | int r; | |
756 | ||
79fac95e | 757 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 758 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 759 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
760 | return r; |
761 | } | |
210c7c4d IE |
762 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
763 | ||
1fe779f8 CO |
764 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
765 | struct | |
766 | kvm_userspace_memory_region *mem, | |
767 | int user_alloc) | |
210c7c4d | 768 | { |
e0d62c7f IE |
769 | if (mem->slot >= KVM_MEMORY_SLOTS) |
770 | return -EINVAL; | |
210c7c4d | 771 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
772 | } |
773 | ||
5bb064dc ZX |
774 | int kvm_get_dirty_log(struct kvm *kvm, |
775 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
776 | { |
777 | struct kvm_memory_slot *memslot; | |
778 | int r, i; | |
87bf6e7d | 779 | unsigned long n; |
6aa8b732 AK |
780 | unsigned long any = 0; |
781 | ||
6aa8b732 AK |
782 | r = -EINVAL; |
783 | if (log->slot >= KVM_MEMORY_SLOTS) | |
784 | goto out; | |
785 | ||
46a26bf5 | 786 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
787 | r = -ENOENT; |
788 | if (!memslot->dirty_bitmap) | |
789 | goto out; | |
790 | ||
87bf6e7d | 791 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 792 | |
cd1a4a98 | 793 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
794 | any = memslot->dirty_bitmap[i]; |
795 | ||
796 | r = -EFAULT; | |
797 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
798 | goto out; | |
799 | ||
5bb064dc ZX |
800 | if (any) |
801 | *is_dirty = 1; | |
6aa8b732 AK |
802 | |
803 | r = 0; | |
6aa8b732 | 804 | out: |
6aa8b732 AK |
805 | return r; |
806 | } | |
807 | ||
54dee993 MT |
808 | void kvm_disable_largepages(void) |
809 | { | |
810 | largepages_enabled = false; | |
811 | } | |
812 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
813 | ||
cea7bb21 IE |
814 | int is_error_page(struct page *page) |
815 | { | |
bf998156 | 816 | return page == bad_page || page == hwpoison_page; |
cea7bb21 IE |
817 | } |
818 | EXPORT_SYMBOL_GPL(is_error_page); | |
819 | ||
35149e21 AL |
820 | int is_error_pfn(pfn_t pfn) |
821 | { | |
bf998156 | 822 | return pfn == bad_pfn || pfn == hwpoison_pfn; |
35149e21 AL |
823 | } |
824 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
825 | ||
bf998156 HY |
826 | int is_hwpoison_pfn(pfn_t pfn) |
827 | { | |
828 | return pfn == hwpoison_pfn; | |
829 | } | |
830 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
831 | ||
f9d46eb0 IE |
832 | static inline unsigned long bad_hva(void) |
833 | { | |
834 | return PAGE_OFFSET; | |
835 | } | |
836 | ||
837 | int kvm_is_error_hva(unsigned long addr) | |
838 | { | |
839 | return addr == bad_hva(); | |
840 | } | |
841 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
842 | ||
2843099f | 843 | struct kvm_memory_slot *gfn_to_memslot_unaliased(struct kvm *kvm, gfn_t gfn) |
6aa8b732 AK |
844 | { |
845 | int i; | |
90d83dc3 | 846 | struct kvm_memslots *slots = kvm_memslots(kvm); |
6aa8b732 | 847 | |
46a26bf5 MT |
848 | for (i = 0; i < slots->nmemslots; ++i) { |
849 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
850 | |
851 | if (gfn >= memslot->base_gfn | |
852 | && gfn < memslot->base_gfn + memslot->npages) | |
853 | return memslot; | |
854 | } | |
8b6d44c7 | 855 | return NULL; |
6aa8b732 | 856 | } |
2843099f | 857 | EXPORT_SYMBOL_GPL(gfn_to_memslot_unaliased); |
e8207547 AK |
858 | |
859 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
860 | { | |
861 | gfn = unalias_gfn(kvm, gfn); | |
2843099f | 862 | return gfn_to_memslot_unaliased(kvm, gfn); |
e8207547 | 863 | } |
6aa8b732 | 864 | |
e0d62c7f IE |
865 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
866 | { | |
867 | int i; | |
90d83dc3 | 868 | struct kvm_memslots *slots = kvm_memslots(kvm); |
e0d62c7f | 869 | |
a983fb23 | 870 | gfn = unalias_gfn_instantiation(kvm, gfn); |
e0d62c7f | 871 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 872 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 873 | |
bc6678a3 MT |
874 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
875 | continue; | |
876 | ||
e0d62c7f IE |
877 | if (gfn >= memslot->base_gfn |
878 | && gfn < memslot->base_gfn + memslot->npages) | |
879 | return 1; | |
880 | } | |
881 | return 0; | |
882 | } | |
883 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
884 | ||
8f0b1ab6 JR |
885 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
886 | { | |
887 | struct vm_area_struct *vma; | |
888 | unsigned long addr, size; | |
889 | ||
890 | size = PAGE_SIZE; | |
891 | ||
892 | addr = gfn_to_hva(kvm, gfn); | |
893 | if (kvm_is_error_hva(addr)) | |
894 | return PAGE_SIZE; | |
895 | ||
896 | down_read(¤t->mm->mmap_sem); | |
897 | vma = find_vma(current->mm, addr); | |
898 | if (!vma) | |
899 | goto out; | |
900 | ||
901 | size = vma_kernel_pagesize(vma); | |
902 | ||
903 | out: | |
904 | up_read(¤t->mm->mmap_sem); | |
905 | ||
906 | return size; | |
907 | } | |
908 | ||
bc6678a3 MT |
909 | int memslot_id(struct kvm *kvm, gfn_t gfn) |
910 | { | |
911 | int i; | |
90d83dc3 | 912 | struct kvm_memslots *slots = kvm_memslots(kvm); |
bc6678a3 MT |
913 | struct kvm_memory_slot *memslot = NULL; |
914 | ||
915 | gfn = unalias_gfn(kvm, gfn); | |
916 | for (i = 0; i < slots->nmemslots; ++i) { | |
917 | memslot = &slots->memslots[i]; | |
918 | ||
919 | if (gfn >= memslot->base_gfn | |
920 | && gfn < memslot->base_gfn + memslot->npages) | |
921 | break; | |
922 | } | |
923 | ||
924 | return memslot - slots->memslots; | |
925 | } | |
926 | ||
f5c98031 TY |
927 | static unsigned long gfn_to_hva_memslot(struct kvm_memory_slot *slot, gfn_t gfn) |
928 | { | |
929 | return slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE; | |
930 | } | |
931 | ||
05da4558 | 932 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) |
539cb660 IE |
933 | { |
934 | struct kvm_memory_slot *slot; | |
935 | ||
a983fb23 | 936 | gfn = unalias_gfn_instantiation(kvm, gfn); |
2843099f | 937 | slot = gfn_to_memslot_unaliased(kvm, gfn); |
bc6678a3 | 938 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 939 | return bad_hva(); |
f5c98031 | 940 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 941 | } |
0d150298 | 942 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 943 | |
506f0d6f | 944 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr) |
954bbbc2 | 945 | { |
8d4e1288 AL |
946 | struct page *page[1]; |
947 | int npages; | |
2e2e3738 | 948 | pfn_t pfn; |
954bbbc2 | 949 | |
60395224 AK |
950 | might_sleep(); |
951 | ||
4c2155ce | 952 | npages = get_user_pages_fast(addr, 1, 1, page); |
539cb660 | 953 | |
2e2e3738 AL |
954 | if (unlikely(npages != 1)) { |
955 | struct vm_area_struct *vma; | |
956 | ||
bf998156 HY |
957 | if (is_hwpoison_address(addr)) { |
958 | get_page(hwpoison_page); | |
959 | return page_to_pfn(hwpoison_page); | |
960 | } | |
961 | ||
4c2155ce | 962 | down_read(¤t->mm->mmap_sem); |
2e2e3738 | 963 | vma = find_vma(current->mm, addr); |
4c2155ce | 964 | |
2e2e3738 AL |
965 | if (vma == NULL || addr < vma->vm_start || |
966 | !(vma->vm_flags & VM_PFNMAP)) { | |
4c2155ce | 967 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
968 | get_page(bad_page); |
969 | return page_to_pfn(bad_page); | |
970 | } | |
971 | ||
972 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
4c2155ce | 973 | up_read(¤t->mm->mmap_sem); |
c77fb9dc | 974 | BUG_ON(!kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
975 | } else |
976 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 977 | |
2e2e3738 | 978 | return pfn; |
35149e21 AL |
979 | } |
980 | ||
506f0d6f MT |
981 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
982 | { | |
983 | unsigned long addr; | |
984 | ||
985 | addr = gfn_to_hva(kvm, gfn); | |
986 | if (kvm_is_error_hva(addr)) { | |
987 | get_page(bad_page); | |
988 | return page_to_pfn(bad_page); | |
989 | } | |
990 | ||
991 | return hva_to_pfn(kvm, addr); | |
992 | } | |
35149e21 AL |
993 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
994 | ||
506f0d6f MT |
995 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
996 | struct kvm_memory_slot *slot, gfn_t gfn) | |
997 | { | |
998 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
999 | return hva_to_pfn(kvm, addr); | |
1000 | } | |
1001 | ||
35149e21 AL |
1002 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1003 | { | |
2e2e3738 AL |
1004 | pfn_t pfn; |
1005 | ||
1006 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1007 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1008 | return pfn_to_page(pfn); |
1009 | ||
c77fb9dc | 1010 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1011 | |
1012 | get_page(bad_page); | |
1013 | return bad_page; | |
954bbbc2 | 1014 | } |
aab61cc0 | 1015 | |
954bbbc2 AK |
1016 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1017 | ||
b4231d61 IE |
1018 | void kvm_release_page_clean(struct page *page) |
1019 | { | |
35149e21 | 1020 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1021 | } |
1022 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1023 | ||
35149e21 AL |
1024 | void kvm_release_pfn_clean(pfn_t pfn) |
1025 | { | |
c77fb9dc | 1026 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1027 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1028 | } |
1029 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1030 | ||
b4231d61 | 1031 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1032 | { |
35149e21 AL |
1033 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1034 | } | |
1035 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1036 | ||
1037 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1038 | { | |
1039 | kvm_set_pfn_dirty(pfn); | |
1040 | kvm_release_pfn_clean(pfn); | |
1041 | } | |
1042 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1043 | ||
1044 | void kvm_set_page_dirty(struct page *page) | |
1045 | { | |
1046 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1047 | } | |
1048 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1049 | ||
1050 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1051 | { | |
c77fb9dc | 1052 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1053 | struct page *page = pfn_to_page(pfn); |
1054 | if (!PageReserved(page)) | |
1055 | SetPageDirty(page); | |
1056 | } | |
8a7ae055 | 1057 | } |
35149e21 AL |
1058 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1059 | ||
1060 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1061 | { | |
c77fb9dc | 1062 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1063 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1064 | } |
1065 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1066 | ||
1067 | void kvm_get_pfn(pfn_t pfn) | |
1068 | { | |
c77fb9dc | 1069 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1070 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1071 | } |
1072 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1073 | |
195aefde IE |
1074 | static int next_segment(unsigned long len, int offset) |
1075 | { | |
1076 | if (len > PAGE_SIZE - offset) | |
1077 | return PAGE_SIZE - offset; | |
1078 | else | |
1079 | return len; | |
1080 | } | |
1081 | ||
1082 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1083 | int len) | |
1084 | { | |
e0506bcb IE |
1085 | int r; |
1086 | unsigned long addr; | |
195aefde | 1087 | |
e0506bcb IE |
1088 | addr = gfn_to_hva(kvm, gfn); |
1089 | if (kvm_is_error_hva(addr)) | |
1090 | return -EFAULT; | |
1091 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1092 | if (r) | |
195aefde | 1093 | return -EFAULT; |
195aefde IE |
1094 | return 0; |
1095 | } | |
1096 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1097 | ||
1098 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1099 | { | |
1100 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1101 | int seg; | |
1102 | int offset = offset_in_page(gpa); | |
1103 | int ret; | |
1104 | ||
1105 | while ((seg = next_segment(len, offset)) != 0) { | |
1106 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1107 | if (ret < 0) | |
1108 | return ret; | |
1109 | offset = 0; | |
1110 | len -= seg; | |
1111 | data += seg; | |
1112 | ++gfn; | |
1113 | } | |
1114 | return 0; | |
1115 | } | |
1116 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1117 | ||
7ec54588 MT |
1118 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1119 | unsigned long len) | |
1120 | { | |
1121 | int r; | |
1122 | unsigned long addr; | |
1123 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1124 | int offset = offset_in_page(gpa); | |
1125 | ||
1126 | addr = gfn_to_hva(kvm, gfn); | |
1127 | if (kvm_is_error_hva(addr)) | |
1128 | return -EFAULT; | |
0aac03f0 | 1129 | pagefault_disable(); |
7ec54588 | 1130 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1131 | pagefault_enable(); |
7ec54588 MT |
1132 | if (r) |
1133 | return -EFAULT; | |
1134 | return 0; | |
1135 | } | |
1136 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1137 | ||
195aefde IE |
1138 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1139 | int offset, int len) | |
1140 | { | |
e0506bcb IE |
1141 | int r; |
1142 | unsigned long addr; | |
195aefde | 1143 | |
e0506bcb IE |
1144 | addr = gfn_to_hva(kvm, gfn); |
1145 | if (kvm_is_error_hva(addr)) | |
1146 | return -EFAULT; | |
1147 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1148 | if (r) | |
195aefde | 1149 | return -EFAULT; |
195aefde IE |
1150 | mark_page_dirty(kvm, gfn); |
1151 | return 0; | |
1152 | } | |
1153 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1154 | ||
1155 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1156 | unsigned long len) | |
1157 | { | |
1158 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1159 | int seg; | |
1160 | int offset = offset_in_page(gpa); | |
1161 | int ret; | |
1162 | ||
1163 | while ((seg = next_segment(len, offset)) != 0) { | |
1164 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1165 | if (ret < 0) | |
1166 | return ret; | |
1167 | offset = 0; | |
1168 | len -= seg; | |
1169 | data += seg; | |
1170 | ++gfn; | |
1171 | } | |
1172 | return 0; | |
1173 | } | |
1174 | ||
1175 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
1176 | { | |
3e021bf5 | 1177 | return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len); |
195aefde IE |
1178 | } |
1179 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1180 | ||
1181 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1182 | { | |
1183 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1184 | int seg; | |
1185 | int offset = offset_in_page(gpa); | |
1186 | int ret; | |
1187 | ||
1188 | while ((seg = next_segment(len, offset)) != 0) { | |
1189 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1190 | if (ret < 0) | |
1191 | return ret; | |
1192 | offset = 0; | |
1193 | len -= seg; | |
1194 | ++gfn; | |
1195 | } | |
1196 | return 0; | |
1197 | } | |
1198 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1199 | ||
6aa8b732 AK |
1200 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1201 | { | |
31389947 | 1202 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1203 | |
3b6fff19 | 1204 | gfn = unalias_gfn(kvm, gfn); |
2843099f | 1205 | memslot = gfn_to_memslot_unaliased(kvm, gfn); |
7e9d619d RR |
1206 | if (memslot && memslot->dirty_bitmap) { |
1207 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1208 | |
d1476937 | 1209 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); |
6aa8b732 AK |
1210 | } |
1211 | } | |
1212 | ||
b6958ce4 ED |
1213 | /* |
1214 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1215 | */ | |
8776e519 | 1216 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1217 | { |
e5c239cf MT |
1218 | DEFINE_WAIT(wait); |
1219 | ||
1220 | for (;;) { | |
1221 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1222 | ||
a1b37100 | 1223 | if (kvm_arch_vcpu_runnable(vcpu)) { |
d7690175 | 1224 | set_bit(KVM_REQ_UNHALT, &vcpu->requests); |
e5c239cf | 1225 | break; |
d7690175 | 1226 | } |
09cec754 GN |
1227 | if (kvm_cpu_has_pending_timer(vcpu)) |
1228 | break; | |
e5c239cf MT |
1229 | if (signal_pending(current)) |
1230 | break; | |
1231 | ||
b6958ce4 | 1232 | schedule(); |
b6958ce4 | 1233 | } |
d3bef15f | 1234 | |
e5c239cf | 1235 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1236 | } |
1237 | ||
6aa8b732 AK |
1238 | void kvm_resched(struct kvm_vcpu *vcpu) |
1239 | { | |
3fca0365 YD |
1240 | if (!need_resched()) |
1241 | return; | |
6aa8b732 | 1242 | cond_resched(); |
6aa8b732 AK |
1243 | } |
1244 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1245 | ||
d255f4f2 ZE |
1246 | void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu) |
1247 | { | |
1248 | ktime_t expires; | |
1249 | DEFINE_WAIT(wait); | |
1250 | ||
1251 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1252 | ||
1253 | /* Sleep for 100 us, and hope lock-holder got scheduled */ | |
1254 | expires = ktime_add_ns(ktime_get(), 100000UL); | |
1255 | schedule_hrtimeout(&expires, HRTIMER_MODE_ABS); | |
1256 | ||
1257 | finish_wait(&vcpu->wq, &wait); | |
1258 | } | |
1259 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1260 | ||
e4a533a4 | 1261 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1262 | { |
1263 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1264 | struct page *page; |
1265 | ||
e4a533a4 | 1266 | if (vmf->pgoff == 0) |
039576c0 | 1267 | page = virt_to_page(vcpu->run); |
09566765 | 1268 | #ifdef CONFIG_X86 |
e4a533a4 | 1269 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1270 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1271 | #endif |
1272 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1273 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1274 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1275 | #endif |
039576c0 | 1276 | else |
e4a533a4 | 1277 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1278 | get_page(page); |
e4a533a4 | 1279 | vmf->page = page; |
1280 | return 0; | |
9a2bb7f4 AK |
1281 | } |
1282 | ||
f0f37e2f | 1283 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1284 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1285 | }; |
1286 | ||
1287 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1288 | { | |
1289 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1290 | return 0; | |
1291 | } | |
1292 | ||
bccf2150 AK |
1293 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1294 | { | |
1295 | struct kvm_vcpu *vcpu = filp->private_data; | |
1296 | ||
66c0b394 | 1297 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1298 | return 0; |
1299 | } | |
1300 | ||
3d3aab1b | 1301 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1302 | .release = kvm_vcpu_release, |
1303 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1304 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1305 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
1306 | }; |
1307 | ||
1308 | /* | |
1309 | * Allocates an inode for the vcpu. | |
1310 | */ | |
1311 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1312 | { | |
628ff7c1 | 1313 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1314 | } |
1315 | ||
c5ea7660 AK |
1316 | /* |
1317 | * Creates some virtual cpus. Good luck creating more than one. | |
1318 | */ | |
73880c80 | 1319 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1320 | { |
1321 | int r; | |
988a2cae | 1322 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1323 | |
73880c80 | 1324 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1325 | if (IS_ERR(vcpu)) |
1326 | return PTR_ERR(vcpu); | |
c5ea7660 | 1327 | |
15ad7146 AK |
1328 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1329 | ||
26e5215f AK |
1330 | r = kvm_arch_vcpu_setup(vcpu); |
1331 | if (r) | |
7d8fece6 | 1332 | return r; |
26e5215f | 1333 | |
11ec2804 | 1334 | mutex_lock(&kvm->lock); |
73880c80 GN |
1335 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1336 | r = -EINVAL; | |
e9b11c17 | 1337 | goto vcpu_destroy; |
fb3f0f51 | 1338 | } |
73880c80 | 1339 | |
988a2cae GN |
1340 | kvm_for_each_vcpu(r, v, kvm) |
1341 | if (v->vcpu_id == id) { | |
73880c80 GN |
1342 | r = -EEXIST; |
1343 | goto vcpu_destroy; | |
1344 | } | |
1345 | ||
1346 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1347 | |
fb3f0f51 | 1348 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1349 | kvm_get_kvm(kvm); |
bccf2150 | 1350 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1351 | if (r < 0) { |
1352 | kvm_put_kvm(kvm); | |
1353 | goto vcpu_destroy; | |
1354 | } | |
1355 | ||
1356 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1357 | smp_wmb(); | |
1358 | atomic_inc(&kvm->online_vcpus); | |
1359 | ||
1360 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1361 | if (kvm->bsp_vcpu_id == id) | |
1362 | kvm->bsp_vcpu = vcpu; | |
1363 | #endif | |
1364 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1365 | return r; |
39c3b86e | 1366 | |
e9b11c17 | 1367 | vcpu_destroy: |
7d8fece6 | 1368 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1369 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1370 | return r; |
1371 | } | |
1372 | ||
1961d276 AK |
1373 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1374 | { | |
1375 | if (sigset) { | |
1376 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1377 | vcpu->sigset_active = 1; | |
1378 | vcpu->sigset = *sigset; | |
1379 | } else | |
1380 | vcpu->sigset_active = 0; | |
1381 | return 0; | |
1382 | } | |
1383 | ||
bccf2150 AK |
1384 | static long kvm_vcpu_ioctl(struct file *filp, |
1385 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1386 | { |
bccf2150 | 1387 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1388 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1389 | int r; |
fa3795a7 DH |
1390 | struct kvm_fpu *fpu = NULL; |
1391 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1392 | |
6d4e4c4f AK |
1393 | if (vcpu->kvm->mm != current->mm) |
1394 | return -EIO; | |
6aa8b732 | 1395 | switch (ioctl) { |
9a2bb7f4 | 1396 | case KVM_RUN: |
f0fe5108 AK |
1397 | r = -EINVAL; |
1398 | if (arg) | |
1399 | goto out; | |
b6c7a5dc | 1400 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 1401 | break; |
6aa8b732 | 1402 | case KVM_GET_REGS: { |
3e4bb3ac | 1403 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1404 | |
3e4bb3ac XZ |
1405 | r = -ENOMEM; |
1406 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1407 | if (!kvm_regs) | |
6aa8b732 | 1408 | goto out; |
3e4bb3ac XZ |
1409 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1410 | if (r) | |
1411 | goto out_free1; | |
6aa8b732 | 1412 | r = -EFAULT; |
3e4bb3ac XZ |
1413 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1414 | goto out_free1; | |
6aa8b732 | 1415 | r = 0; |
3e4bb3ac XZ |
1416 | out_free1: |
1417 | kfree(kvm_regs); | |
6aa8b732 AK |
1418 | break; |
1419 | } | |
1420 | case KVM_SET_REGS: { | |
3e4bb3ac | 1421 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1422 | |
3e4bb3ac XZ |
1423 | r = -ENOMEM; |
1424 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1425 | if (!kvm_regs) | |
6aa8b732 | 1426 | goto out; |
3e4bb3ac XZ |
1427 | r = -EFAULT; |
1428 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1429 | goto out_free2; | |
1430 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1431 | if (r) |
3e4bb3ac | 1432 | goto out_free2; |
6aa8b732 | 1433 | r = 0; |
3e4bb3ac XZ |
1434 | out_free2: |
1435 | kfree(kvm_regs); | |
6aa8b732 AK |
1436 | break; |
1437 | } | |
1438 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1439 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1440 | r = -ENOMEM; | |
1441 | if (!kvm_sregs) | |
1442 | goto out; | |
1443 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1444 | if (r) |
1445 | goto out; | |
1446 | r = -EFAULT; | |
fa3795a7 | 1447 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1448 | goto out; |
1449 | r = 0; | |
1450 | break; | |
1451 | } | |
1452 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1453 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1454 | r = -ENOMEM; | |
1455 | if (!kvm_sregs) | |
1456 | goto out; | |
6aa8b732 | 1457 | r = -EFAULT; |
fa3795a7 | 1458 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1459 | goto out; |
fa3795a7 | 1460 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1461 | if (r) |
1462 | goto out; | |
1463 | r = 0; | |
1464 | break; | |
1465 | } | |
62d9f0db MT |
1466 | case KVM_GET_MP_STATE: { |
1467 | struct kvm_mp_state mp_state; | |
1468 | ||
1469 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1470 | if (r) | |
1471 | goto out; | |
1472 | r = -EFAULT; | |
1473 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1474 | goto out; | |
1475 | r = 0; | |
1476 | break; | |
1477 | } | |
1478 | case KVM_SET_MP_STATE: { | |
1479 | struct kvm_mp_state mp_state; | |
1480 | ||
1481 | r = -EFAULT; | |
1482 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1483 | goto out; | |
1484 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1485 | if (r) | |
1486 | goto out; | |
1487 | r = 0; | |
1488 | break; | |
1489 | } | |
6aa8b732 AK |
1490 | case KVM_TRANSLATE: { |
1491 | struct kvm_translation tr; | |
1492 | ||
1493 | r = -EFAULT; | |
2f366987 | 1494 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1495 | goto out; |
8b006791 | 1496 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1497 | if (r) |
1498 | goto out; | |
1499 | r = -EFAULT; | |
2f366987 | 1500 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1501 | goto out; |
1502 | r = 0; | |
1503 | break; | |
1504 | } | |
d0bfb940 JK |
1505 | case KVM_SET_GUEST_DEBUG: { |
1506 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1507 | |
1508 | r = -EFAULT; | |
2f366987 | 1509 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1510 | goto out; |
d0bfb940 | 1511 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1512 | if (r) |
1513 | goto out; | |
1514 | r = 0; | |
1515 | break; | |
1516 | } | |
1961d276 AK |
1517 | case KVM_SET_SIGNAL_MASK: { |
1518 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1519 | struct kvm_signal_mask kvm_sigmask; | |
1520 | sigset_t sigset, *p; | |
1521 | ||
1522 | p = NULL; | |
1523 | if (argp) { | |
1524 | r = -EFAULT; | |
1525 | if (copy_from_user(&kvm_sigmask, argp, | |
1526 | sizeof kvm_sigmask)) | |
1527 | goto out; | |
1528 | r = -EINVAL; | |
1529 | if (kvm_sigmask.len != sizeof sigset) | |
1530 | goto out; | |
1531 | r = -EFAULT; | |
1532 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1533 | sizeof sigset)) | |
1534 | goto out; | |
1535 | p = &sigset; | |
1536 | } | |
1537 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1538 | break; | |
1539 | } | |
b8836737 | 1540 | case KVM_GET_FPU: { |
fa3795a7 DH |
1541 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1542 | r = -ENOMEM; | |
1543 | if (!fpu) | |
1544 | goto out; | |
1545 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1546 | if (r) |
1547 | goto out; | |
1548 | r = -EFAULT; | |
fa3795a7 | 1549 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1550 | goto out; |
1551 | r = 0; | |
1552 | break; | |
1553 | } | |
1554 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1555 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1556 | r = -ENOMEM; | |
1557 | if (!fpu) | |
1558 | goto out; | |
b8836737 | 1559 | r = -EFAULT; |
fa3795a7 | 1560 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1561 | goto out; |
fa3795a7 | 1562 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1563 | if (r) |
1564 | goto out; | |
1565 | r = 0; | |
1566 | break; | |
1567 | } | |
bccf2150 | 1568 | default: |
313a3dc7 | 1569 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1570 | } |
1571 | out: | |
fa3795a7 DH |
1572 | kfree(fpu); |
1573 | kfree(kvm_sregs); | |
bccf2150 AK |
1574 | return r; |
1575 | } | |
1576 | ||
1577 | static long kvm_vm_ioctl(struct file *filp, | |
1578 | unsigned int ioctl, unsigned long arg) | |
1579 | { | |
1580 | struct kvm *kvm = filp->private_data; | |
1581 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1582 | int r; |
bccf2150 | 1583 | |
6d4e4c4f AK |
1584 | if (kvm->mm != current->mm) |
1585 | return -EIO; | |
bccf2150 AK |
1586 | switch (ioctl) { |
1587 | case KVM_CREATE_VCPU: | |
1588 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1589 | if (r < 0) | |
1590 | goto out; | |
1591 | break; | |
6fc138d2 IE |
1592 | case KVM_SET_USER_MEMORY_REGION: { |
1593 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1594 | ||
1595 | r = -EFAULT; | |
1596 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1597 | sizeof kvm_userspace_mem)) | |
1598 | goto out; | |
1599 | ||
1600 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1601 | if (r) |
1602 | goto out; | |
1603 | break; | |
1604 | } | |
1605 | case KVM_GET_DIRTY_LOG: { | |
1606 | struct kvm_dirty_log log; | |
1607 | ||
1608 | r = -EFAULT; | |
2f366987 | 1609 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1610 | goto out; |
2c6f5df9 | 1611 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1612 | if (r) |
1613 | goto out; | |
1614 | break; | |
1615 | } | |
5f94c174 LV |
1616 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1617 | case KVM_REGISTER_COALESCED_MMIO: { | |
1618 | struct kvm_coalesced_mmio_zone zone; | |
1619 | r = -EFAULT; | |
1620 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1621 | goto out; | |
5f94c174 LV |
1622 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
1623 | if (r) | |
1624 | goto out; | |
1625 | r = 0; | |
1626 | break; | |
1627 | } | |
1628 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1629 | struct kvm_coalesced_mmio_zone zone; | |
1630 | r = -EFAULT; | |
1631 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1632 | goto out; | |
5f94c174 LV |
1633 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
1634 | if (r) | |
1635 | goto out; | |
1636 | r = 0; | |
1637 | break; | |
1638 | } | |
1639 | #endif | |
721eecbf GH |
1640 | case KVM_IRQFD: { |
1641 | struct kvm_irqfd data; | |
1642 | ||
1643 | r = -EFAULT; | |
1644 | if (copy_from_user(&data, argp, sizeof data)) | |
1645 | goto out; | |
1646 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1647 | break; | |
1648 | } | |
d34e6b17 GH |
1649 | case KVM_IOEVENTFD: { |
1650 | struct kvm_ioeventfd data; | |
1651 | ||
1652 | r = -EFAULT; | |
1653 | if (copy_from_user(&data, argp, sizeof data)) | |
1654 | goto out; | |
1655 | r = kvm_ioeventfd(kvm, &data); | |
1656 | break; | |
1657 | } | |
73880c80 GN |
1658 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1659 | case KVM_SET_BOOT_CPU_ID: | |
1660 | r = 0; | |
894a9c55 | 1661 | mutex_lock(&kvm->lock); |
73880c80 GN |
1662 | if (atomic_read(&kvm->online_vcpus) != 0) |
1663 | r = -EBUSY; | |
1664 | else | |
1665 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1666 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1667 | break; |
1668 | #endif | |
f17abe9a | 1669 | default: |
1fe779f8 | 1670 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1671 | if (r == -ENOTTY) |
1672 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1673 | } |
1674 | out: | |
1675 | return r; | |
1676 | } | |
1677 | ||
6ff5894c AB |
1678 | #ifdef CONFIG_COMPAT |
1679 | struct compat_kvm_dirty_log { | |
1680 | __u32 slot; | |
1681 | __u32 padding1; | |
1682 | union { | |
1683 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1684 | __u64 padding2; | |
1685 | }; | |
1686 | }; | |
1687 | ||
1688 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1689 | unsigned int ioctl, unsigned long arg) | |
1690 | { | |
1691 | struct kvm *kvm = filp->private_data; | |
1692 | int r; | |
1693 | ||
1694 | if (kvm->mm != current->mm) | |
1695 | return -EIO; | |
1696 | switch (ioctl) { | |
1697 | case KVM_GET_DIRTY_LOG: { | |
1698 | struct compat_kvm_dirty_log compat_log; | |
1699 | struct kvm_dirty_log log; | |
1700 | ||
1701 | r = -EFAULT; | |
1702 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1703 | sizeof(compat_log))) | |
1704 | goto out; | |
1705 | log.slot = compat_log.slot; | |
1706 | log.padding1 = compat_log.padding1; | |
1707 | log.padding2 = compat_log.padding2; | |
1708 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
1709 | ||
1710 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
1711 | if (r) | |
1712 | goto out; | |
1713 | break; | |
1714 | } | |
1715 | default: | |
1716 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
1717 | } | |
1718 | ||
1719 | out: | |
1720 | return r; | |
1721 | } | |
1722 | #endif | |
1723 | ||
e4a533a4 | 1724 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 1725 | { |
777b3f49 MT |
1726 | struct page *page[1]; |
1727 | unsigned long addr; | |
1728 | int npages; | |
1729 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 1730 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 1731 | |
777b3f49 MT |
1732 | addr = gfn_to_hva(kvm, gfn); |
1733 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 1734 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1735 | |
1736 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
1737 | NULL); | |
1738 | if (unlikely(npages != 1)) | |
e4a533a4 | 1739 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
1740 | |
1741 | vmf->page = page[0]; | |
e4a533a4 | 1742 | return 0; |
f17abe9a AK |
1743 | } |
1744 | ||
f0f37e2f | 1745 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 1746 | .fault = kvm_vm_fault, |
f17abe9a AK |
1747 | }; |
1748 | ||
1749 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
1750 | { | |
1751 | vma->vm_ops = &kvm_vm_vm_ops; | |
1752 | return 0; | |
1753 | } | |
1754 | ||
3d3aab1b | 1755 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
1756 | .release = kvm_vm_release, |
1757 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
1758 | #ifdef CONFIG_COMPAT |
1759 | .compat_ioctl = kvm_vm_compat_ioctl, | |
1760 | #endif | |
f17abe9a AK |
1761 | .mmap = kvm_vm_mmap, |
1762 | }; | |
1763 | ||
1764 | static int kvm_dev_ioctl_create_vm(void) | |
1765 | { | |
6ce5a090 | 1766 | int fd, r; |
f17abe9a AK |
1767 | struct kvm *kvm; |
1768 | ||
f17abe9a | 1769 | kvm = kvm_create_vm(); |
d6d28168 AK |
1770 | if (IS_ERR(kvm)) |
1771 | return PTR_ERR(kvm); | |
6ce5a090 TY |
1772 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1773 | r = kvm_coalesced_mmio_init(kvm); | |
1774 | if (r < 0) { | |
1775 | kvm_put_kvm(kvm); | |
1776 | return r; | |
1777 | } | |
1778 | #endif | |
628ff7c1 | 1779 | fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2030a42c | 1780 | if (fd < 0) |
66c0b394 | 1781 | kvm_put_kvm(kvm); |
f17abe9a | 1782 | |
f17abe9a | 1783 | return fd; |
f17abe9a AK |
1784 | } |
1785 | ||
1a811b61 AK |
1786 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
1787 | { | |
1788 | switch (arg) { | |
ca9edaee | 1789 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 1790 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 1791 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
1792 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1793 | case KVM_CAP_SET_BOOT_CPU_ID: | |
1794 | #endif | |
a9c7399d | 1795 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 1796 | return 1; |
399ec807 AK |
1797 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
1798 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 1799 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 1800 | #endif |
1a811b61 AK |
1801 | default: |
1802 | break; | |
1803 | } | |
1804 | return kvm_dev_ioctl_check_extension(arg); | |
1805 | } | |
1806 | ||
f17abe9a AK |
1807 | static long kvm_dev_ioctl(struct file *filp, |
1808 | unsigned int ioctl, unsigned long arg) | |
1809 | { | |
07c45a36 | 1810 | long r = -EINVAL; |
f17abe9a AK |
1811 | |
1812 | switch (ioctl) { | |
1813 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
1814 | r = -EINVAL; |
1815 | if (arg) | |
1816 | goto out; | |
f17abe9a AK |
1817 | r = KVM_API_VERSION; |
1818 | break; | |
1819 | case KVM_CREATE_VM: | |
f0fe5108 AK |
1820 | r = -EINVAL; |
1821 | if (arg) | |
1822 | goto out; | |
f17abe9a AK |
1823 | r = kvm_dev_ioctl_create_vm(); |
1824 | break; | |
018d00d2 | 1825 | case KVM_CHECK_EXTENSION: |
1a811b61 | 1826 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 1827 | break; |
07c45a36 AK |
1828 | case KVM_GET_VCPU_MMAP_SIZE: |
1829 | r = -EINVAL; | |
1830 | if (arg) | |
1831 | goto out; | |
adb1ff46 AK |
1832 | r = PAGE_SIZE; /* struct kvm_run */ |
1833 | #ifdef CONFIG_X86 | |
1834 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
1835 | #endif |
1836 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1837 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 1838 | #endif |
07c45a36 | 1839 | break; |
d4c9ff2d FEL |
1840 | case KVM_TRACE_ENABLE: |
1841 | case KVM_TRACE_PAUSE: | |
1842 | case KVM_TRACE_DISABLE: | |
2023a29c | 1843 | r = -EOPNOTSUPP; |
d4c9ff2d | 1844 | break; |
6aa8b732 | 1845 | default: |
043405e1 | 1846 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
1847 | } |
1848 | out: | |
1849 | return r; | |
1850 | } | |
1851 | ||
6aa8b732 | 1852 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
1853 | .unlocked_ioctl = kvm_dev_ioctl, |
1854 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
1855 | }; |
1856 | ||
1857 | static struct miscdevice kvm_dev = { | |
bbe4432e | 1858 | KVM_MINOR, |
6aa8b732 AK |
1859 | "kvm", |
1860 | &kvm_chardev_ops, | |
1861 | }; | |
1862 | ||
1b6c0168 AK |
1863 | static void hardware_enable(void *junk) |
1864 | { | |
1865 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 1866 | int r; |
1b6c0168 | 1867 | |
7f59f492 | 1868 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1869 | return; |
10474ae8 | 1870 | |
7f59f492 | 1871 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
1872 | |
1873 | r = kvm_arch_hardware_enable(NULL); | |
1874 | ||
1875 | if (r) { | |
1876 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
1877 | atomic_inc(&hardware_enable_failed); | |
1878 | printk(KERN_INFO "kvm: enabling virtualization on " | |
1879 | "CPU%d failed\n", cpu); | |
1880 | } | |
1b6c0168 AK |
1881 | } |
1882 | ||
1883 | static void hardware_disable(void *junk) | |
1884 | { | |
1885 | int cpu = raw_smp_processor_id(); | |
1886 | ||
7f59f492 | 1887 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 1888 | return; |
7f59f492 | 1889 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 1890 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
1891 | } |
1892 | ||
10474ae8 AG |
1893 | static void hardware_disable_all_nolock(void) |
1894 | { | |
1895 | BUG_ON(!kvm_usage_count); | |
1896 | ||
1897 | kvm_usage_count--; | |
1898 | if (!kvm_usage_count) | |
1899 | on_each_cpu(hardware_disable, NULL, 1); | |
1900 | } | |
1901 | ||
1902 | static void hardware_disable_all(void) | |
1903 | { | |
1904 | spin_lock(&kvm_lock); | |
1905 | hardware_disable_all_nolock(); | |
1906 | spin_unlock(&kvm_lock); | |
1907 | } | |
1908 | ||
1909 | static int hardware_enable_all(void) | |
1910 | { | |
1911 | int r = 0; | |
1912 | ||
1913 | spin_lock(&kvm_lock); | |
1914 | ||
1915 | kvm_usage_count++; | |
1916 | if (kvm_usage_count == 1) { | |
1917 | atomic_set(&hardware_enable_failed, 0); | |
1918 | on_each_cpu(hardware_enable, NULL, 1); | |
1919 | ||
1920 | if (atomic_read(&hardware_enable_failed)) { | |
1921 | hardware_disable_all_nolock(); | |
1922 | r = -EBUSY; | |
1923 | } | |
1924 | } | |
1925 | ||
1926 | spin_unlock(&kvm_lock); | |
1927 | ||
1928 | return r; | |
1929 | } | |
1930 | ||
774c47f1 AK |
1931 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
1932 | void *v) | |
1933 | { | |
1934 | int cpu = (long)v; | |
1935 | ||
10474ae8 AG |
1936 | if (!kvm_usage_count) |
1937 | return NOTIFY_OK; | |
1938 | ||
1a6f4d7f | 1939 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 1940 | switch (val) { |
cec9ad27 | 1941 | case CPU_DYING: |
6ec8a856 AK |
1942 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
1943 | cpu); | |
1944 | hardware_disable(NULL); | |
1945 | break; | |
43934a38 JK |
1946 | case CPU_ONLINE: |
1947 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", | |
1948 | cpu); | |
8691e5a8 | 1949 | smp_call_function_single(cpu, hardware_enable, NULL, 1); |
774c47f1 AK |
1950 | break; |
1951 | } | |
1952 | return NOTIFY_OK; | |
1953 | } | |
1954 | ||
4ecac3fd AK |
1955 | |
1956 | asmlinkage void kvm_handle_fault_on_reboot(void) | |
1957 | { | |
1958 | if (kvm_rebooting) | |
1959 | /* spin while reset goes on */ | |
1960 | while (true) | |
1961 | ; | |
1962 | /* Fault while not rebooting. We want the trace. */ | |
1963 | BUG(); | |
1964 | } | |
1965 | EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot); | |
1966 | ||
9a2b85c6 | 1967 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 1968 | void *v) |
9a2b85c6 | 1969 | { |
8e1c1815 SY |
1970 | /* |
1971 | * Some (well, at least mine) BIOSes hang on reboot if | |
1972 | * in vmx root mode. | |
1973 | * | |
1974 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
1975 | */ | |
1976 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
1977 | kvm_rebooting = true; | |
1978 | on_each_cpu(hardware_disable, NULL, 1); | |
9a2b85c6 RR |
1979 | return NOTIFY_OK; |
1980 | } | |
1981 | ||
1982 | static struct notifier_block kvm_reboot_notifier = { | |
1983 | .notifier_call = kvm_reboot, | |
1984 | .priority = 0, | |
1985 | }; | |
1986 | ||
e93f8a0f | 1987 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
1988 | { |
1989 | int i; | |
1990 | ||
1991 | for (i = 0; i < bus->dev_count; i++) { | |
1992 | struct kvm_io_device *pos = bus->devs[i]; | |
1993 | ||
1994 | kvm_iodevice_destructor(pos); | |
1995 | } | |
e93f8a0f | 1996 | kfree(bus); |
2eeb2e94 GH |
1997 | } |
1998 | ||
bda9020e | 1999 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2000 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2001 | int len, const void *val) |
2eeb2e94 GH |
2002 | { |
2003 | int i; | |
90d83dc3 LJ |
2004 | struct kvm_io_bus *bus; |
2005 | ||
2006 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
bda9020e MT |
2007 | for (i = 0; i < bus->dev_count; i++) |
2008 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
2009 | return 0; | |
2010 | return -EOPNOTSUPP; | |
2011 | } | |
2eeb2e94 | 2012 | |
bda9020e | 2013 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2014 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2015 | int len, void *val) | |
bda9020e MT |
2016 | { |
2017 | int i; | |
90d83dc3 | 2018 | struct kvm_io_bus *bus; |
e93f8a0f | 2019 | |
90d83dc3 | 2020 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
bda9020e MT |
2021 | for (i = 0; i < bus->dev_count; i++) |
2022 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
2023 | return 0; | |
2024 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
2025 | } |
2026 | ||
79fac95e | 2027 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2028 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2029 | struct kvm_io_device *dev) | |
6c474694 | 2030 | { |
e93f8a0f | 2031 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2032 | |
e93f8a0f | 2033 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2034 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2035 | return -ENOSPC; | |
2eeb2e94 | 2036 | |
e93f8a0f MT |
2037 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2038 | if (!new_bus) | |
2039 | return -ENOMEM; | |
2040 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2041 | new_bus->devs[new_bus->dev_count++] = dev; | |
2042 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2043 | synchronize_srcu_expedited(&kvm->srcu); | |
2044 | kfree(bus); | |
090b7aff GH |
2045 | |
2046 | return 0; | |
2047 | } | |
2048 | ||
79fac95e | 2049 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2050 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2051 | struct kvm_io_device *dev) | |
090b7aff | 2052 | { |
e93f8a0f MT |
2053 | int i, r; |
2054 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2055 | |
e93f8a0f MT |
2056 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2057 | if (!new_bus) | |
2058 | return -ENOMEM; | |
090b7aff | 2059 | |
e93f8a0f MT |
2060 | bus = kvm->buses[bus_idx]; |
2061 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2062 | ||
2063 | r = -ENOENT; | |
2064 | for (i = 0; i < new_bus->dev_count; i++) | |
2065 | if (new_bus->devs[i] == dev) { | |
2066 | r = 0; | |
2067 | new_bus->devs[i] = new_bus->devs[--new_bus->dev_count]; | |
090b7aff GH |
2068 | break; |
2069 | } | |
e93f8a0f MT |
2070 | |
2071 | if (r) { | |
2072 | kfree(new_bus); | |
2073 | return r; | |
2074 | } | |
2075 | ||
2076 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2077 | synchronize_srcu_expedited(&kvm->srcu); | |
2078 | kfree(bus); | |
2079 | return r; | |
2eeb2e94 GH |
2080 | } |
2081 | ||
774c47f1 AK |
2082 | static struct notifier_block kvm_cpu_notifier = { |
2083 | .notifier_call = kvm_cpu_hotplug, | |
2084 | .priority = 20, /* must be > scheduler priority */ | |
2085 | }; | |
2086 | ||
8b88b099 | 2087 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2088 | { |
2089 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2090 | struct kvm *kvm; |
2091 | ||
8b88b099 | 2092 | *val = 0; |
ba1389b7 AK |
2093 | spin_lock(&kvm_lock); |
2094 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2095 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2096 | spin_unlock(&kvm_lock); |
8b88b099 | 2097 | return 0; |
ba1389b7 AK |
2098 | } |
2099 | ||
2100 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2101 | ||
8b88b099 | 2102 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2103 | { |
2104 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2105 | struct kvm *kvm; |
2106 | struct kvm_vcpu *vcpu; | |
2107 | int i; | |
2108 | ||
8b88b099 | 2109 | *val = 0; |
1165f5fe AK |
2110 | spin_lock(&kvm_lock); |
2111 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
2112 | kvm_for_each_vcpu(i, vcpu, kvm) |
2113 | *val += *(u32 *)((void *)vcpu + offset); | |
2114 | ||
1165f5fe | 2115 | spin_unlock(&kvm_lock); |
8b88b099 | 2116 | return 0; |
1165f5fe AK |
2117 | } |
2118 | ||
ba1389b7 AK |
2119 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2120 | ||
828c0950 | 2121 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2122 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2123 | [KVM_STAT_VM] = &vm_stat_fops, | |
2124 | }; | |
1165f5fe | 2125 | |
a16b043c | 2126 | static void kvm_init_debug(void) |
6aa8b732 AK |
2127 | { |
2128 | struct kvm_stats_debugfs_item *p; | |
2129 | ||
76f7c879 | 2130 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2131 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2132 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2133 | (void *)(long)p->offset, |
ba1389b7 | 2134 | stat_fops[p->kind]); |
6aa8b732 AK |
2135 | } |
2136 | ||
2137 | static void kvm_exit_debug(void) | |
2138 | { | |
2139 | struct kvm_stats_debugfs_item *p; | |
2140 | ||
2141 | for (p = debugfs_entries; p->name; ++p) | |
2142 | debugfs_remove(p->dentry); | |
76f7c879 | 2143 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2144 | } |
2145 | ||
59ae6c6b AK |
2146 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2147 | { | |
10474ae8 AG |
2148 | if (kvm_usage_count) |
2149 | hardware_disable(NULL); | |
59ae6c6b AK |
2150 | return 0; |
2151 | } | |
2152 | ||
2153 | static int kvm_resume(struct sys_device *dev) | |
2154 | { | |
10474ae8 AG |
2155 | if (kvm_usage_count) |
2156 | hardware_enable(NULL); | |
59ae6c6b AK |
2157 | return 0; |
2158 | } | |
2159 | ||
2160 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2161 | .name = "kvm", |
59ae6c6b AK |
2162 | .suspend = kvm_suspend, |
2163 | .resume = kvm_resume, | |
2164 | }; | |
2165 | ||
2166 | static struct sys_device kvm_sysdev = { | |
2167 | .id = 0, | |
2168 | .cls = &kvm_sysdev_class, | |
2169 | }; | |
2170 | ||
cea7bb21 | 2171 | struct page *bad_page; |
35149e21 | 2172 | pfn_t bad_pfn; |
6aa8b732 | 2173 | |
15ad7146 AK |
2174 | static inline |
2175 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2176 | { | |
2177 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2178 | } | |
2179 | ||
2180 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2181 | { | |
2182 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2183 | ||
e9b11c17 | 2184 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2185 | } |
2186 | ||
2187 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2188 | struct task_struct *next) | |
2189 | { | |
2190 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2191 | ||
e9b11c17 | 2192 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2193 | } |
2194 | ||
0ee75bea | 2195 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2196 | struct module *module) |
6aa8b732 AK |
2197 | { |
2198 | int r; | |
002c7f7c | 2199 | int cpu; |
6aa8b732 | 2200 | |
f8c16bba ZX |
2201 | r = kvm_arch_init(opaque); |
2202 | if (r) | |
d2308784 | 2203 | goto out_fail; |
cb498ea2 ZX |
2204 | |
2205 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2206 | ||
2207 | if (bad_page == NULL) { | |
2208 | r = -ENOMEM; | |
2209 | goto out; | |
2210 | } | |
2211 | ||
35149e21 AL |
2212 | bad_pfn = page_to_pfn(bad_page); |
2213 | ||
bf998156 HY |
2214 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2215 | ||
2216 | if (hwpoison_page == NULL) { | |
2217 | r = -ENOMEM; | |
2218 | goto out_free_0; | |
2219 | } | |
2220 | ||
2221 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2222 | ||
8437a617 | 2223 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2224 | r = -ENOMEM; |
2225 | goto out_free_0; | |
2226 | } | |
2227 | ||
e9b11c17 | 2228 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2229 | if (r < 0) |
7f59f492 | 2230 | goto out_free_0a; |
6aa8b732 | 2231 | |
002c7f7c YS |
2232 | for_each_online_cpu(cpu) { |
2233 | smp_call_function_single(cpu, | |
e9b11c17 | 2234 | kvm_arch_check_processor_compat, |
8691e5a8 | 2235 | &r, 1); |
002c7f7c | 2236 | if (r < 0) |
d2308784 | 2237 | goto out_free_1; |
002c7f7c YS |
2238 | } |
2239 | ||
774c47f1 AK |
2240 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2241 | if (r) | |
d2308784 | 2242 | goto out_free_2; |
6aa8b732 AK |
2243 | register_reboot_notifier(&kvm_reboot_notifier); |
2244 | ||
59ae6c6b AK |
2245 | r = sysdev_class_register(&kvm_sysdev_class); |
2246 | if (r) | |
d2308784 | 2247 | goto out_free_3; |
59ae6c6b AK |
2248 | |
2249 | r = sysdev_register(&kvm_sysdev); | |
2250 | if (r) | |
d2308784 | 2251 | goto out_free_4; |
59ae6c6b | 2252 | |
c16f862d | 2253 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2254 | if (!vcpu_align) |
2255 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2256 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2257 | 0, NULL); |
c16f862d RR |
2258 | if (!kvm_vcpu_cache) { |
2259 | r = -ENOMEM; | |
d2308784 | 2260 | goto out_free_5; |
c16f862d RR |
2261 | } |
2262 | ||
6aa8b732 | 2263 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2264 | kvm_vm_fops.owner = module; |
2265 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2266 | |
2267 | r = misc_register(&kvm_dev); | |
2268 | if (r) { | |
d77c26fc | 2269 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
2270 | goto out_free; |
2271 | } | |
2272 | ||
15ad7146 AK |
2273 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2274 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2275 | ||
0ea4ed8e DW |
2276 | kvm_init_debug(); |
2277 | ||
c7addb90 | 2278 | return 0; |
6aa8b732 AK |
2279 | |
2280 | out_free: | |
c16f862d | 2281 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2282 | out_free_5: |
59ae6c6b | 2283 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2284 | out_free_4: |
59ae6c6b | 2285 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2286 | out_free_3: |
6aa8b732 | 2287 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2288 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2289 | out_free_2: |
d2308784 | 2290 | out_free_1: |
e9b11c17 | 2291 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2292 | out_free_0a: |
2293 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2294 | out_free_0: |
bf998156 HY |
2295 | if (hwpoison_page) |
2296 | __free_page(hwpoison_page); | |
d2308784 | 2297 | __free_page(bad_page); |
ca45aaae | 2298 | out: |
f8c16bba | 2299 | kvm_arch_exit(); |
d2308784 | 2300 | out_fail: |
6aa8b732 AK |
2301 | return r; |
2302 | } | |
cb498ea2 | 2303 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2304 | |
cb498ea2 | 2305 | void kvm_exit(void) |
6aa8b732 | 2306 | { |
0ea4ed8e | 2307 | kvm_exit_debug(); |
6aa8b732 | 2308 | misc_deregister(&kvm_dev); |
c16f862d | 2309 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
2310 | sysdev_unregister(&kvm_sysdev); |
2311 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2312 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2313 | unregister_cpu_notifier(&kvm_cpu_notifier); |
15c8b6c1 | 2314 | on_each_cpu(hardware_disable, NULL, 1); |
e9b11c17 | 2315 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2316 | kvm_arch_exit(); |
7f59f492 | 2317 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2318 | __free_page(hwpoison_page); |
cea7bb21 | 2319 | __free_page(bad_page); |
6aa8b732 | 2320 | } |
cb498ea2 | 2321 | EXPORT_SYMBOL_GPL(kvm_exit); |