]>
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> | |
fb3600cc | 33 | #include <linux/syscore_ops.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> |
743eeb0b SL |
50 | #include <linux/sort.h> |
51 | #include <linux/bsearch.h> | |
6aa8b732 | 52 | |
e495606d | 53 | #include <asm/processor.h> |
e495606d AK |
54 | #include <asm/io.h> |
55 | #include <asm/uaccess.h> | |
3e021bf5 | 56 | #include <asm/pgtable.h> |
6aa8b732 | 57 | |
5f94c174 | 58 | #include "coalesced_mmio.h" |
af585b92 | 59 | #include "async_pf.h" |
5f94c174 | 60 | |
229456fc MT |
61 | #define CREATE_TRACE_POINTS |
62 | #include <trace/events/kvm.h> | |
63 | ||
6aa8b732 AK |
64 | MODULE_AUTHOR("Qumranet"); |
65 | MODULE_LICENSE("GPL"); | |
66 | ||
fa40a821 MT |
67 | /* |
68 | * Ordering of locks: | |
69 | * | |
fae3a353 | 70 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
71 | */ |
72 | ||
e935b837 | 73 | DEFINE_RAW_SPINLOCK(kvm_lock); |
e9b11c17 | 74 | LIST_HEAD(vm_list); |
133de902 | 75 | |
7f59f492 | 76 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
77 | static int kvm_usage_count = 0; |
78 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 79 | |
c16f862d RR |
80 | struct kmem_cache *kvm_vcpu_cache; |
81 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 82 | |
15ad7146 AK |
83 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
84 | ||
76f7c879 | 85 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 86 | |
bccf2150 AK |
87 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
88 | unsigned long arg); | |
1dda606c AG |
89 | #ifdef CONFIG_COMPAT |
90 | static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl, | |
91 | unsigned long arg); | |
92 | #endif | |
10474ae8 AG |
93 | static int hardware_enable_all(void); |
94 | static void hardware_disable_all(void); | |
bccf2150 | 95 | |
e93f8a0f MT |
96 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
97 | ||
b7c4145b AK |
98 | bool kvm_rebooting; |
99 | EXPORT_SYMBOL_GPL(kvm_rebooting); | |
4ecac3fd | 100 | |
54dee993 MT |
101 | static bool largepages_enabled = true; |
102 | ||
fa7bff8f GN |
103 | static struct page *hwpoison_page; |
104 | static pfn_t hwpoison_pfn; | |
bf998156 | 105 | |
fce92dce XG |
106 | struct page *fault_page; |
107 | pfn_t fault_pfn; | |
edba23e5 | 108 | |
c77fb9dc | 109 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 110 | { |
fc5659c8 | 111 | if (pfn_valid(pfn)) { |
22e5c47e | 112 | int reserved; |
936a5fe6 | 113 | struct page *tail = pfn_to_page(pfn); |
22e5c47e AA |
114 | struct page *head = compound_trans_head(tail); |
115 | reserved = PageReserved(head); | |
936a5fe6 | 116 | if (head != tail) { |
936a5fe6 | 117 | /* |
22e5c47e AA |
118 | * "head" is not a dangling pointer |
119 | * (compound_trans_head takes care of that) | |
120 | * but the hugepage may have been splitted | |
121 | * from under us (and we may not hold a | |
122 | * reference count on the head page so it can | |
123 | * be reused before we run PageReferenced), so | |
124 | * we've to check PageTail before returning | |
125 | * what we just read. | |
936a5fe6 | 126 | */ |
22e5c47e AA |
127 | smp_rmb(); |
128 | if (PageTail(tail)) | |
129 | return reserved; | |
936a5fe6 AA |
130 | } |
131 | return PageReserved(tail); | |
fc5659c8 | 132 | } |
cbff90a7 BAY |
133 | |
134 | return true; | |
135 | } | |
136 | ||
bccf2150 AK |
137 | /* |
138 | * Switches to specified vcpu, until a matching vcpu_put() | |
139 | */ | |
313a3dc7 | 140 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 141 | { |
15ad7146 AK |
142 | int cpu; |
143 | ||
bccf2150 | 144 | mutex_lock(&vcpu->mutex); |
34bb10b7 RR |
145 | if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) { |
146 | /* The thread running this VCPU changed. */ | |
147 | struct pid *oldpid = vcpu->pid; | |
148 | struct pid *newpid = get_task_pid(current, PIDTYPE_PID); | |
149 | rcu_assign_pointer(vcpu->pid, newpid); | |
150 | synchronize_rcu(); | |
151 | put_pid(oldpid); | |
152 | } | |
15ad7146 AK |
153 | cpu = get_cpu(); |
154 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 155 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 156 | put_cpu(); |
6aa8b732 AK |
157 | } |
158 | ||
313a3dc7 | 159 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 160 | { |
15ad7146 | 161 | preempt_disable(); |
313a3dc7 | 162 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
163 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
164 | preempt_enable(); | |
6aa8b732 AK |
165 | mutex_unlock(&vcpu->mutex); |
166 | } | |
167 | ||
d9e368d6 AK |
168 | static void ack_flush(void *_completed) |
169 | { | |
d9e368d6 AK |
170 | } |
171 | ||
49846896 | 172 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 173 | { |
597a5f55 | 174 | int i, cpu, me; |
6ef7a1bc RR |
175 | cpumask_var_t cpus; |
176 | bool called = true; | |
d9e368d6 | 177 | struct kvm_vcpu *vcpu; |
d9e368d6 | 178 | |
79f55997 | 179 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 180 | |
3cba4130 | 181 | me = get_cpu(); |
988a2cae | 182 | kvm_for_each_vcpu(i, vcpu, kvm) { |
3cba4130 | 183 | kvm_make_request(req, vcpu); |
d9e368d6 | 184 | cpu = vcpu->cpu; |
6b7e2d09 XG |
185 | |
186 | /* Set ->requests bit before we read ->mode */ | |
187 | smp_mb(); | |
188 | ||
189 | if (cpus != NULL && cpu != -1 && cpu != me && | |
190 | kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE) | |
6ef7a1bc | 191 | cpumask_set_cpu(cpu, cpus); |
49846896 | 192 | } |
6ef7a1bc RR |
193 | if (unlikely(cpus == NULL)) |
194 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
195 | else if (!cpumask_empty(cpus)) | |
196 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
197 | else | |
198 | called = false; | |
3cba4130 | 199 | put_cpu(); |
6ef7a1bc | 200 | free_cpumask_var(cpus); |
49846896 | 201 | return called; |
d9e368d6 AK |
202 | } |
203 | ||
49846896 | 204 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 205 | { |
a4ee1ca4 XG |
206 | int dirty_count = kvm->tlbs_dirty; |
207 | ||
208 | smp_mb(); | |
49846896 RR |
209 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
210 | ++kvm->stat.remote_tlb_flush; | |
a4ee1ca4 | 211 | cmpxchg(&kvm->tlbs_dirty, dirty_count, 0); |
2e53d63a MT |
212 | } |
213 | ||
49846896 RR |
214 | void kvm_reload_remote_mmus(struct kvm *kvm) |
215 | { | |
216 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
217 | } | |
2e53d63a | 218 | |
fb3f0f51 RR |
219 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
220 | { | |
221 | struct page *page; | |
222 | int r; | |
223 | ||
224 | mutex_init(&vcpu->mutex); | |
225 | vcpu->cpu = -1; | |
fb3f0f51 RR |
226 | vcpu->kvm = kvm; |
227 | vcpu->vcpu_id = id; | |
34bb10b7 | 228 | vcpu->pid = NULL; |
b6958ce4 | 229 | init_waitqueue_head(&vcpu->wq); |
af585b92 | 230 | kvm_async_pf_vcpu_init(vcpu); |
fb3f0f51 RR |
231 | |
232 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
233 | if (!page) { | |
234 | r = -ENOMEM; | |
235 | goto fail; | |
236 | } | |
237 | vcpu->run = page_address(page); | |
238 | ||
e9b11c17 | 239 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 240 | if (r < 0) |
e9b11c17 | 241 | goto fail_free_run; |
fb3f0f51 RR |
242 | return 0; |
243 | ||
fb3f0f51 RR |
244 | fail_free_run: |
245 | free_page((unsigned long)vcpu->run); | |
246 | fail: | |
76fafa5e | 247 | return r; |
fb3f0f51 RR |
248 | } |
249 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
250 | ||
251 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
252 | { | |
34bb10b7 | 253 | put_pid(vcpu->pid); |
e9b11c17 | 254 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
255 | free_page((unsigned long)vcpu->run); |
256 | } | |
257 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
258 | ||
e930bffe AA |
259 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
260 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
261 | { | |
262 | return container_of(mn, struct kvm, mmu_notifier); | |
263 | } | |
264 | ||
265 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
266 | struct mm_struct *mm, | |
267 | unsigned long address) | |
268 | { | |
269 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 270 | int need_tlb_flush, idx; |
e930bffe AA |
271 | |
272 | /* | |
273 | * When ->invalidate_page runs, the linux pte has been zapped | |
274 | * already but the page is still allocated until | |
275 | * ->invalidate_page returns. So if we increase the sequence | |
276 | * here the kvm page fault will notice if the spte can't be | |
277 | * established because the page is going to be freed. If | |
278 | * instead the kvm page fault establishes the spte before | |
279 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
280 | * before returning. | |
281 | * | |
282 | * The sequence increase only need to be seen at spin_unlock | |
283 | * time, and not at spin_lock time. | |
284 | * | |
285 | * Increasing the sequence after the spin_unlock would be | |
286 | * unsafe because the kvm page fault could then establish the | |
287 | * pte after kvm_unmap_hva returned, without noticing the page | |
288 | * is going to be freed. | |
289 | */ | |
bc6678a3 | 290 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
291 | spin_lock(&kvm->mmu_lock); |
292 | kvm->mmu_notifier_seq++; | |
a4ee1ca4 | 293 | need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty; |
e930bffe | 294 | spin_unlock(&kvm->mmu_lock); |
bc6678a3 | 295 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
296 | |
297 | /* we've to flush the tlb before the pages can be freed */ | |
298 | if (need_tlb_flush) | |
299 | kvm_flush_remote_tlbs(kvm); | |
300 | ||
301 | } | |
302 | ||
3da0dd43 IE |
303 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
304 | struct mm_struct *mm, | |
305 | unsigned long address, | |
306 | pte_t pte) | |
307 | { | |
308 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 309 | int idx; |
3da0dd43 | 310 | |
bc6678a3 | 311 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
312 | spin_lock(&kvm->mmu_lock); |
313 | kvm->mmu_notifier_seq++; | |
314 | kvm_set_spte_hva(kvm, address, pte); | |
315 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 316 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
317 | } |
318 | ||
e930bffe AA |
319 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
320 | struct mm_struct *mm, | |
321 | unsigned long start, | |
322 | unsigned long end) | |
323 | { | |
324 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 325 | int need_tlb_flush = 0, idx; |
e930bffe | 326 | |
bc6678a3 | 327 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
328 | spin_lock(&kvm->mmu_lock); |
329 | /* | |
330 | * The count increase must become visible at unlock time as no | |
331 | * spte can be established without taking the mmu_lock and | |
332 | * count is also read inside the mmu_lock critical section. | |
333 | */ | |
334 | kvm->mmu_notifier_count++; | |
335 | for (; start < end; start += PAGE_SIZE) | |
336 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
a4ee1ca4 | 337 | need_tlb_flush |= kvm->tlbs_dirty; |
e930bffe | 338 | spin_unlock(&kvm->mmu_lock); |
bc6678a3 | 339 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
340 | |
341 | /* we've to flush the tlb before the pages can be freed */ | |
342 | if (need_tlb_flush) | |
343 | kvm_flush_remote_tlbs(kvm); | |
344 | } | |
345 | ||
346 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
347 | struct mm_struct *mm, | |
348 | unsigned long start, | |
349 | unsigned long end) | |
350 | { | |
351 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
352 | ||
353 | spin_lock(&kvm->mmu_lock); | |
354 | /* | |
355 | * This sequence increase will notify the kvm page fault that | |
356 | * the page that is going to be mapped in the spte could have | |
357 | * been freed. | |
358 | */ | |
359 | kvm->mmu_notifier_seq++; | |
360 | /* | |
361 | * The above sequence increase must be visible before the | |
362 | * below count decrease but both values are read by the kvm | |
363 | * page fault under mmu_lock spinlock so we don't need to add | |
364 | * a smb_wmb() here in between the two. | |
365 | */ | |
366 | kvm->mmu_notifier_count--; | |
367 | spin_unlock(&kvm->mmu_lock); | |
368 | ||
369 | BUG_ON(kvm->mmu_notifier_count < 0); | |
370 | } | |
371 | ||
372 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
373 | struct mm_struct *mm, | |
374 | unsigned long address) | |
375 | { | |
376 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 377 | int young, idx; |
e930bffe | 378 | |
bc6678a3 | 379 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
380 | spin_lock(&kvm->mmu_lock); |
381 | young = kvm_age_hva(kvm, address); | |
382 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 383 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
384 | |
385 | if (young) | |
386 | kvm_flush_remote_tlbs(kvm); | |
387 | ||
388 | return young; | |
389 | } | |
390 | ||
8ee53820 AA |
391 | static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn, |
392 | struct mm_struct *mm, | |
393 | unsigned long address) | |
394 | { | |
395 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
396 | int young, idx; | |
397 | ||
398 | idx = srcu_read_lock(&kvm->srcu); | |
399 | spin_lock(&kvm->mmu_lock); | |
400 | young = kvm_test_age_hva(kvm, address); | |
401 | spin_unlock(&kvm->mmu_lock); | |
402 | srcu_read_unlock(&kvm->srcu, idx); | |
403 | ||
404 | return young; | |
405 | } | |
406 | ||
85db06e5 MT |
407 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
408 | struct mm_struct *mm) | |
409 | { | |
410 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
411 | int idx; |
412 | ||
413 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 414 | kvm_arch_flush_shadow(kvm); |
eda2beda | 415 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
416 | } |
417 | ||
e930bffe AA |
418 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
419 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
420 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
421 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
422 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
8ee53820 | 423 | .test_young = kvm_mmu_notifier_test_young, |
3da0dd43 | 424 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 425 | .release = kvm_mmu_notifier_release, |
e930bffe | 426 | }; |
4c07b0a4 AK |
427 | |
428 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
429 | { | |
430 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
431 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
432 | } | |
433 | ||
434 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
435 | ||
436 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
437 | { | |
438 | return 0; | |
439 | } | |
440 | ||
e930bffe AA |
441 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
442 | ||
bf3e05bc XG |
443 | static void kvm_init_memslots_id(struct kvm *kvm) |
444 | { | |
445 | int i; | |
446 | struct kvm_memslots *slots = kvm->memslots; | |
447 | ||
448 | for (i = 0; i < KVM_MEM_SLOTS_NUM; i++) | |
449 | slots->memslots[i].id = i; | |
450 | } | |
451 | ||
f17abe9a | 452 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 453 | { |
d89f5eff JK |
454 | int r, i; |
455 | struct kvm *kvm = kvm_arch_alloc_vm(); | |
6aa8b732 | 456 | |
d89f5eff JK |
457 | if (!kvm) |
458 | return ERR_PTR(-ENOMEM); | |
459 | ||
460 | r = kvm_arch_init_vm(kvm); | |
461 | if (r) | |
462 | goto out_err_nodisable; | |
10474ae8 AG |
463 | |
464 | r = hardware_enable_all(); | |
465 | if (r) | |
466 | goto out_err_nodisable; | |
467 | ||
75858a84 AK |
468 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
469 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 470 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 471 | #endif |
6aa8b732 | 472 | |
46a26bf5 MT |
473 | r = -ENOMEM; |
474 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
475 | if (!kvm->memslots) | |
57e7fbee | 476 | goto out_err_nosrcu; |
bf3e05bc | 477 | kvm_init_memslots_id(kvm); |
bc6678a3 | 478 | if (init_srcu_struct(&kvm->srcu)) |
57e7fbee | 479 | goto out_err_nosrcu; |
e93f8a0f MT |
480 | for (i = 0; i < KVM_NR_BUSES; i++) { |
481 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
482 | GFP_KERNEL); | |
57e7fbee | 483 | if (!kvm->buses[i]) |
e93f8a0f | 484 | goto out_err; |
e93f8a0f | 485 | } |
e930bffe | 486 | |
74b5c5bf | 487 | spin_lock_init(&kvm->mmu_lock); |
6d4e4c4f AK |
488 | kvm->mm = current->mm; |
489 | atomic_inc(&kvm->mm->mm_count); | |
d34e6b17 | 490 | kvm_eventfd_init(kvm); |
11ec2804 | 491 | mutex_init(&kvm->lock); |
60eead79 | 492 | mutex_init(&kvm->irq_lock); |
79fac95e | 493 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 494 | atomic_set(&kvm->users_count, 1); |
74b5c5bf MW |
495 | |
496 | r = kvm_init_mmu_notifier(kvm); | |
497 | if (r) | |
498 | goto out_err; | |
499 | ||
e935b837 | 500 | raw_spin_lock(&kvm_lock); |
5e58cfe4 | 501 | list_add(&kvm->vm_list, &vm_list); |
e935b837 | 502 | raw_spin_unlock(&kvm_lock); |
d89f5eff | 503 | |
f17abe9a | 504 | return kvm; |
10474ae8 AG |
505 | |
506 | out_err: | |
57e7fbee JK |
507 | cleanup_srcu_struct(&kvm->srcu); |
508 | out_err_nosrcu: | |
10474ae8 AG |
509 | hardware_disable_all(); |
510 | out_err_nodisable: | |
e93f8a0f MT |
511 | for (i = 0; i < KVM_NR_BUSES; i++) |
512 | kfree(kvm->buses[i]); | |
46a26bf5 | 513 | kfree(kvm->memslots); |
d89f5eff | 514 | kvm_arch_free_vm(kvm); |
10474ae8 | 515 | return ERR_PTR(r); |
f17abe9a AK |
516 | } |
517 | ||
a36a57b1 TY |
518 | static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot) |
519 | { | |
520 | if (!memslot->dirty_bitmap) | |
521 | return; | |
522 | ||
6f9e5c17 TY |
523 | if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE) |
524 | vfree(memslot->dirty_bitmap_head); | |
525 | else | |
526 | kfree(memslot->dirty_bitmap_head); | |
527 | ||
a36a57b1 | 528 | memslot->dirty_bitmap = NULL; |
515a0127 | 529 | memslot->dirty_bitmap_head = NULL; |
a36a57b1 TY |
530 | } |
531 | ||
6aa8b732 AK |
532 | /* |
533 | * Free any memory in @free but not in @dont. | |
534 | */ | |
535 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
536 | struct kvm_memory_slot *dont) | |
537 | { | |
ec04b260 JR |
538 | int i; |
539 | ||
290fc38d IE |
540 | if (!dont || free->rmap != dont->rmap) |
541 | vfree(free->rmap); | |
6aa8b732 AK |
542 | |
543 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
a36a57b1 | 544 | kvm_destroy_dirty_bitmap(free); |
6aa8b732 | 545 | |
ec04b260 JR |
546 | |
547 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
548 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
549 | vfree(free->lpage_info[i]); | |
550 | free->lpage_info[i] = NULL; | |
551 | } | |
552 | } | |
05da4558 | 553 | |
6aa8b732 | 554 | free->npages = 0; |
8d4e1288 | 555 | free->rmap = NULL; |
6aa8b732 AK |
556 | } |
557 | ||
d19a9cd2 | 558 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 | 559 | { |
46a26bf5 | 560 | struct kvm_memslots *slots = kvm->memslots; |
be6ba0f0 | 561 | struct kvm_memory_slot *memslot; |
46a26bf5 | 562 | |
be6ba0f0 XG |
563 | kvm_for_each_memslot(memslot, slots) |
564 | kvm_free_physmem_slot(memslot, NULL); | |
6aa8b732 | 565 | |
46a26bf5 | 566 | kfree(kvm->memslots); |
6aa8b732 AK |
567 | } |
568 | ||
f17abe9a AK |
569 | static void kvm_destroy_vm(struct kvm *kvm) |
570 | { | |
e93f8a0f | 571 | int i; |
6d4e4c4f AK |
572 | struct mm_struct *mm = kvm->mm; |
573 | ||
ad8ba2cd | 574 | kvm_arch_sync_events(kvm); |
e935b837 | 575 | raw_spin_lock(&kvm_lock); |
133de902 | 576 | list_del(&kvm->vm_list); |
e935b837 | 577 | raw_spin_unlock(&kvm_lock); |
399ec807 | 578 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
579 | for (i = 0; i < KVM_NR_BUSES; i++) |
580 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 581 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
582 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
583 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
584 | #else |
585 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 586 | #endif |
d19a9cd2 | 587 | kvm_arch_destroy_vm(kvm); |
d89f5eff JK |
588 | kvm_free_physmem(kvm); |
589 | cleanup_srcu_struct(&kvm->srcu); | |
590 | kvm_arch_free_vm(kvm); | |
10474ae8 | 591 | hardware_disable_all(); |
6d4e4c4f | 592 | mmdrop(mm); |
f17abe9a AK |
593 | } |
594 | ||
d39f13b0 IE |
595 | void kvm_get_kvm(struct kvm *kvm) |
596 | { | |
597 | atomic_inc(&kvm->users_count); | |
598 | } | |
599 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
600 | ||
601 | void kvm_put_kvm(struct kvm *kvm) | |
602 | { | |
603 | if (atomic_dec_and_test(&kvm->users_count)) | |
604 | kvm_destroy_vm(kvm); | |
605 | } | |
606 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
607 | ||
608 | ||
f17abe9a AK |
609 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
610 | { | |
611 | struct kvm *kvm = filp->private_data; | |
612 | ||
721eecbf GH |
613 | kvm_irqfd_release(kvm); |
614 | ||
d39f13b0 | 615 | kvm_put_kvm(kvm); |
6aa8b732 AK |
616 | return 0; |
617 | } | |
618 | ||
d48ead8b | 619 | #ifndef CONFIG_S390 |
515a0127 TY |
620 | /* |
621 | * Allocation size is twice as large as the actual dirty bitmap size. | |
622 | * This makes it possible to do double buffering: see x86's | |
623 | * kvm_vm_ioctl_get_dirty_log(). | |
624 | */ | |
a36a57b1 TY |
625 | static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot) |
626 | { | |
515a0127 | 627 | unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot); |
a36a57b1 | 628 | |
6f9e5c17 TY |
629 | if (dirty_bytes > PAGE_SIZE) |
630 | memslot->dirty_bitmap = vzalloc(dirty_bytes); | |
631 | else | |
632 | memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL); | |
633 | ||
a36a57b1 TY |
634 | if (!memslot->dirty_bitmap) |
635 | return -ENOMEM; | |
636 | ||
515a0127 | 637 | memslot->dirty_bitmap_head = memslot->dirty_bitmap; |
7850ac54 | 638 | memslot->nr_dirty_pages = 0; |
a36a57b1 TY |
639 | return 0; |
640 | } | |
d48ead8b | 641 | #endif /* !CONFIG_S390 */ |
a36a57b1 | 642 | |
bf3e05bc XG |
643 | static struct kvm_memory_slot * |
644 | search_memslots(struct kvm_memslots *slots, gfn_t gfn) | |
645 | { | |
646 | struct kvm_memory_slot *memslot; | |
647 | ||
648 | kvm_for_each_memslot(memslot, slots) | |
649 | if (gfn >= memslot->base_gfn && | |
650 | gfn < memslot->base_gfn + memslot->npages) | |
651 | return memslot; | |
652 | ||
653 | return NULL; | |
654 | } | |
655 | ||
656 | static int cmp_memslot(const void *slot1, const void *slot2) | |
657 | { | |
658 | struct kvm_memory_slot *s1, *s2; | |
659 | ||
660 | s1 = (struct kvm_memory_slot *)slot1; | |
661 | s2 = (struct kvm_memory_slot *)slot2; | |
662 | ||
663 | if (s1->npages < s2->npages) | |
664 | return 1; | |
665 | if (s1->npages > s2->npages) | |
666 | return -1; | |
667 | ||
668 | return 0; | |
669 | } | |
670 | ||
671 | /* | |
672 | * Sort the memslots base on its size, so the larger slots | |
673 | * will get better fit. | |
674 | */ | |
675 | static void sort_memslots(struct kvm_memslots *slots) | |
676 | { | |
677 | sort(slots->memslots, KVM_MEM_SLOTS_NUM, | |
678 | sizeof(struct kvm_memory_slot), cmp_memslot, NULL); | |
679 | } | |
680 | ||
be593d62 XG |
681 | void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new) |
682 | { | |
683 | if (new) { | |
684 | int id = new->id; | |
28a37544 | 685 | struct kvm_memory_slot *old = id_to_memslot(slots, id); |
bf3e05bc | 686 | unsigned long npages = old->npages; |
be593d62 | 687 | |
28a37544 | 688 | *old = *new; |
bf3e05bc XG |
689 | if (new->npages != npages) |
690 | sort_memslots(slots); | |
be593d62 XG |
691 | } |
692 | ||
693 | slots->generation++; | |
694 | } | |
695 | ||
6aa8b732 AK |
696 | /* |
697 | * Allocate some memory and give it an address in the guest physical address | |
698 | * space. | |
699 | * | |
700 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 701 | * |
10589a46 | 702 | * Must be called holding mmap_sem for write. |
6aa8b732 | 703 | */ |
f78e0e2e SY |
704 | int __kvm_set_memory_region(struct kvm *kvm, |
705 | struct kvm_userspace_memory_region *mem, | |
706 | int user_alloc) | |
6aa8b732 | 707 | { |
8234b22e | 708 | int r; |
6aa8b732 | 709 | gfn_t base_gfn; |
28bcb112 HC |
710 | unsigned long npages; |
711 | unsigned long i; | |
6aa8b732 AK |
712 | struct kvm_memory_slot *memslot; |
713 | struct kvm_memory_slot old, new; | |
bc6678a3 | 714 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
715 | |
716 | r = -EINVAL; | |
717 | /* General sanity checks */ | |
718 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
719 | goto out; | |
720 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
721 | goto out; | |
fa3d315a TY |
722 | /* We can read the guest memory with __xxx_user() later on. */ |
723 | if (user_alloc && | |
724 | ((mem->userspace_addr & (PAGE_SIZE - 1)) || | |
9e3bb6b6 HC |
725 | !access_ok(VERIFY_WRITE, |
726 | (void __user *)(unsigned long)mem->userspace_addr, | |
727 | mem->memory_size))) | |
78749809 | 728 | goto out; |
93a5cef0 | 729 | if (mem->slot >= KVM_MEM_SLOTS_NUM) |
6aa8b732 AK |
730 | goto out; |
731 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
732 | goto out; | |
733 | ||
28a37544 | 734 | memslot = id_to_memslot(kvm->memslots, mem->slot); |
6aa8b732 AK |
735 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
736 | npages = mem->memory_size >> PAGE_SHIFT; | |
737 | ||
660c22c4 TY |
738 | r = -EINVAL; |
739 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
740 | goto out; | |
741 | ||
6aa8b732 AK |
742 | if (!npages) |
743 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
744 | ||
6aa8b732 AK |
745 | new = old = *memslot; |
746 | ||
e36d96f7 | 747 | new.id = mem->slot; |
6aa8b732 AK |
748 | new.base_gfn = base_gfn; |
749 | new.npages = npages; | |
750 | new.flags = mem->flags; | |
751 | ||
752 | /* Disallow changing a memory slot's size. */ | |
753 | r = -EINVAL; | |
754 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 755 | goto out_free; |
6aa8b732 AK |
756 | |
757 | /* Check for overlaps */ | |
758 | r = -EEXIST; | |
759 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 760 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 761 | |
4cd481f6 | 762 | if (s == memslot || !s->npages) |
6aa8b732 AK |
763 | continue; |
764 | if (!((base_gfn + npages <= s->base_gfn) || | |
765 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 766 | goto out_free; |
6aa8b732 | 767 | } |
6aa8b732 | 768 | |
6aa8b732 AK |
769 | /* Free page dirty bitmap if unneeded */ |
770 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 771 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
772 | |
773 | r = -ENOMEM; | |
774 | ||
775 | /* Allocate if a slot is being created */ | |
eff0114a | 776 | #ifndef CONFIG_S390 |
8d4e1288 | 777 | if (npages && !new.rmap) { |
26535037 | 778 | new.rmap = vzalloc(npages * sizeof(*new.rmap)); |
290fc38d IE |
779 | |
780 | if (!new.rmap) | |
f78e0e2e | 781 | goto out_free; |
290fc38d | 782 | |
80b14b5b | 783 | new.user_alloc = user_alloc; |
bc6678a3 | 784 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 785 | } |
ec04b260 JR |
786 | if (!npages) |
787 | goto skip_lpage; | |
05da4558 | 788 | |
ec04b260 | 789 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
790 | unsigned long ugfn; |
791 | unsigned long j; | |
792 | int lpages; | |
ec04b260 | 793 | int level = i + 2; |
05da4558 | 794 | |
ec04b260 JR |
795 | /* Avoid unused variable warning if no large pages */ |
796 | (void)level; | |
797 | ||
798 | if (new.lpage_info[i]) | |
799 | continue; | |
800 | ||
82855413 JR |
801 | lpages = 1 + ((base_gfn + npages - 1) |
802 | >> KVM_HPAGE_GFN_SHIFT(level)); | |
803 | lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level); | |
ec04b260 | 804 | |
26535037 | 805 | new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i])); |
ec04b260 JR |
806 | |
807 | if (!new.lpage_info[i]) | |
05da4558 MT |
808 | goto out_free; |
809 | ||
82855413 | 810 | if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 811 | new.lpage_info[i][0].write_count = 1; |
82855413 | 812 | if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 813 | new.lpage_info[i][lpages - 1].write_count = 1; |
ac04527f AK |
814 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
815 | /* | |
816 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
817 | * other, or if explicitly asked to, disable large page |
818 | * support for this slot | |
ac04527f | 819 | */ |
ec04b260 | 820 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 821 | !largepages_enabled) |
ec04b260 JR |
822 | for (j = 0; j < lpages; ++j) |
823 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 824 | } |
6aa8b732 | 825 | |
ec04b260 JR |
826 | skip_lpage: |
827 | ||
6aa8b732 AK |
828 | /* Allocate page dirty bitmap if needed */ |
829 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
a36a57b1 | 830 | if (kvm_create_dirty_bitmap(&new) < 0) |
f78e0e2e | 831 | goto out_free; |
bc6678a3 | 832 | /* destroy any largepage mappings for dirty tracking */ |
6aa8b732 | 833 | } |
3eea8437 CB |
834 | #else /* not defined CONFIG_S390 */ |
835 | new.user_alloc = user_alloc; | |
836 | if (user_alloc) | |
837 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 838 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 839 | |
bc6678a3 | 840 | if (!npages) { |
28a37544 XG |
841 | struct kvm_memory_slot *slot; |
842 | ||
bc6678a3 | 843 | r = -ENOMEM; |
6da64fdb TM |
844 | slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots), |
845 | GFP_KERNEL); | |
bc6678a3 MT |
846 | if (!slots) |
847 | goto out_free; | |
28a37544 XG |
848 | slot = id_to_memslot(slots, mem->slot); |
849 | slot->flags |= KVM_MEMSLOT_INVALID; | |
850 | ||
be593d62 | 851 | update_memslots(slots, NULL); |
bc6678a3 MT |
852 | |
853 | old_memslots = kvm->memslots; | |
854 | rcu_assign_pointer(kvm->memslots, slots); | |
855 | synchronize_srcu_expedited(&kvm->srcu); | |
856 | /* From this point no new shadow pages pointing to a deleted | |
857 | * memslot will be created. | |
858 | * | |
859 | * validation of sp->gfn happens in: | |
860 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
861 | * - kvm_is_visible_gfn (mmu_check_roots) | |
862 | */ | |
34d4cb8f | 863 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
864 | kfree(old_memslots); |
865 | } | |
34d4cb8f | 866 | |
f7784b8e MT |
867 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
868 | if (r) | |
869 | goto out_free; | |
870 | ||
bc6678a3 MT |
871 | /* map the pages in iommu page table */ |
872 | if (npages) { | |
873 | r = kvm_iommu_map_pages(kvm, &new); | |
874 | if (r) | |
875 | goto out_free; | |
876 | } | |
604b38ac | 877 | |
bc6678a3 | 878 | r = -ENOMEM; |
6da64fdb TM |
879 | slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots), |
880 | GFP_KERNEL); | |
bc6678a3 MT |
881 | if (!slots) |
882 | goto out_free; | |
bc6678a3 MT |
883 | |
884 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
885 | if (!npages) { | |
886 | new.rmap = NULL; | |
887 | new.dirty_bitmap = NULL; | |
888 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
889 | new.lpage_info[i] = NULL; | |
890 | } | |
891 | ||
be593d62 | 892 | update_memslots(slots, &new); |
bc6678a3 MT |
893 | old_memslots = kvm->memslots; |
894 | rcu_assign_pointer(kvm->memslots, slots); | |
895 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 896 | |
f7784b8e | 897 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 898 | |
ce88decf XG |
899 | /* |
900 | * If the new memory slot is created, we need to clear all | |
901 | * mmio sptes. | |
902 | */ | |
903 | if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT) | |
904 | kvm_arch_flush_shadow(kvm); | |
905 | ||
bc6678a3 MT |
906 | kvm_free_physmem_slot(&old, &new); |
907 | kfree(old_memslots); | |
908 | ||
6aa8b732 AK |
909 | return 0; |
910 | ||
f78e0e2e | 911 | out_free: |
6aa8b732 AK |
912 | kvm_free_physmem_slot(&new, &old); |
913 | out: | |
914 | return r; | |
210c7c4d IE |
915 | |
916 | } | |
f78e0e2e SY |
917 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
918 | ||
919 | int kvm_set_memory_region(struct kvm *kvm, | |
920 | struct kvm_userspace_memory_region *mem, | |
921 | int user_alloc) | |
922 | { | |
923 | int r; | |
924 | ||
79fac95e | 925 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 926 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 927 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
928 | return r; |
929 | } | |
210c7c4d IE |
930 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
931 | ||
1fe779f8 CO |
932 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
933 | struct | |
934 | kvm_userspace_memory_region *mem, | |
935 | int user_alloc) | |
210c7c4d | 936 | { |
e0d62c7f IE |
937 | if (mem->slot >= KVM_MEMORY_SLOTS) |
938 | return -EINVAL; | |
210c7c4d | 939 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
940 | } |
941 | ||
5bb064dc ZX |
942 | int kvm_get_dirty_log(struct kvm *kvm, |
943 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
944 | { |
945 | struct kvm_memory_slot *memslot; | |
946 | int r, i; | |
87bf6e7d | 947 | unsigned long n; |
6aa8b732 AK |
948 | unsigned long any = 0; |
949 | ||
6aa8b732 AK |
950 | r = -EINVAL; |
951 | if (log->slot >= KVM_MEMORY_SLOTS) | |
952 | goto out; | |
953 | ||
28a37544 | 954 | memslot = id_to_memslot(kvm->memslots, log->slot); |
6aa8b732 AK |
955 | r = -ENOENT; |
956 | if (!memslot->dirty_bitmap) | |
957 | goto out; | |
958 | ||
87bf6e7d | 959 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 960 | |
cd1a4a98 | 961 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
962 | any = memslot->dirty_bitmap[i]; |
963 | ||
964 | r = -EFAULT; | |
965 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
966 | goto out; | |
967 | ||
5bb064dc ZX |
968 | if (any) |
969 | *is_dirty = 1; | |
6aa8b732 AK |
970 | |
971 | r = 0; | |
6aa8b732 | 972 | out: |
6aa8b732 AK |
973 | return r; |
974 | } | |
975 | ||
54dee993 MT |
976 | void kvm_disable_largepages(void) |
977 | { | |
978 | largepages_enabled = false; | |
979 | } | |
980 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
981 | ||
cea7bb21 IE |
982 | int is_error_page(struct page *page) |
983 | { | |
edba23e5 | 984 | return page == bad_page || page == hwpoison_page || page == fault_page; |
cea7bb21 IE |
985 | } |
986 | EXPORT_SYMBOL_GPL(is_error_page); | |
987 | ||
35149e21 AL |
988 | int is_error_pfn(pfn_t pfn) |
989 | { | |
edba23e5 | 990 | return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn; |
35149e21 AL |
991 | } |
992 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
993 | ||
bf998156 HY |
994 | int is_hwpoison_pfn(pfn_t pfn) |
995 | { | |
996 | return pfn == hwpoison_pfn; | |
997 | } | |
998 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
999 | ||
edba23e5 GN |
1000 | int is_fault_pfn(pfn_t pfn) |
1001 | { | |
1002 | return pfn == fault_pfn; | |
1003 | } | |
1004 | EXPORT_SYMBOL_GPL(is_fault_pfn); | |
1005 | ||
fce92dce XG |
1006 | int is_noslot_pfn(pfn_t pfn) |
1007 | { | |
1008 | return pfn == bad_pfn; | |
1009 | } | |
1010 | EXPORT_SYMBOL_GPL(is_noslot_pfn); | |
1011 | ||
1012 | int is_invalid_pfn(pfn_t pfn) | |
1013 | { | |
1014 | return pfn == hwpoison_pfn || pfn == fault_pfn; | |
1015 | } | |
1016 | EXPORT_SYMBOL_GPL(is_invalid_pfn); | |
1017 | ||
f9d46eb0 IE |
1018 | static inline unsigned long bad_hva(void) |
1019 | { | |
1020 | return PAGE_OFFSET; | |
1021 | } | |
1022 | ||
1023 | int kvm_is_error_hva(unsigned long addr) | |
1024 | { | |
1025 | return addr == bad_hva(); | |
1026 | } | |
1027 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
1028 | ||
49c7754c GN |
1029 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots, |
1030 | gfn_t gfn) | |
6aa8b732 | 1031 | { |
bf3e05bc | 1032 | return search_memslots(slots, gfn); |
6aa8b732 | 1033 | } |
49c7754c GN |
1034 | |
1035 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
1036 | { | |
1037 | return __gfn_to_memslot(kvm_memslots(kvm), gfn); | |
1038 | } | |
a1f4d395 | 1039 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 1040 | |
e0d62c7f IE |
1041 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
1042 | { | |
bf3e05bc | 1043 | struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn); |
e0d62c7f | 1044 | |
bf3e05bc XG |
1045 | if (!memslot || memslot->id >= KVM_MEMORY_SLOTS || |
1046 | memslot->flags & KVM_MEMSLOT_INVALID) | |
1047 | return 0; | |
e0d62c7f | 1048 | |
bf3e05bc | 1049 | return 1; |
e0d62c7f IE |
1050 | } |
1051 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
1052 | ||
8f0b1ab6 JR |
1053 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
1054 | { | |
1055 | struct vm_area_struct *vma; | |
1056 | unsigned long addr, size; | |
1057 | ||
1058 | size = PAGE_SIZE; | |
1059 | ||
1060 | addr = gfn_to_hva(kvm, gfn); | |
1061 | if (kvm_is_error_hva(addr)) | |
1062 | return PAGE_SIZE; | |
1063 | ||
1064 | down_read(¤t->mm->mmap_sem); | |
1065 | vma = find_vma(current->mm, addr); | |
1066 | if (!vma) | |
1067 | goto out; | |
1068 | ||
1069 | size = vma_kernel_pagesize(vma); | |
1070 | ||
1071 | out: | |
1072 | up_read(¤t->mm->mmap_sem); | |
1073 | ||
1074 | return size; | |
1075 | } | |
1076 | ||
49c7754c | 1077 | static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn, |
48987781 | 1078 | gfn_t *nr_pages) |
539cb660 | 1079 | { |
bc6678a3 | 1080 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 1081 | return bad_hva(); |
48987781 XG |
1082 | |
1083 | if (nr_pages) | |
1084 | *nr_pages = slot->npages - (gfn - slot->base_gfn); | |
1085 | ||
f5c98031 | 1086 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 1087 | } |
48987781 XG |
1088 | |
1089 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) | |
1090 | { | |
49c7754c | 1091 | return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL); |
48987781 | 1092 | } |
0d150298 | 1093 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1094 | |
8030089f GN |
1095 | static pfn_t get_fault_pfn(void) |
1096 | { | |
1097 | get_page(fault_page); | |
1098 | return fault_pfn; | |
1099 | } | |
1100 | ||
0857b9e9 GN |
1101 | int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm, |
1102 | unsigned long start, int write, struct page **page) | |
1103 | { | |
1104 | int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET; | |
1105 | ||
1106 | if (write) | |
1107 | flags |= FOLL_WRITE; | |
1108 | ||
1109 | return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL); | |
1110 | } | |
1111 | ||
fafc3dba HY |
1112 | static inline int check_user_page_hwpoison(unsigned long addr) |
1113 | { | |
1114 | int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE; | |
1115 | ||
1116 | rc = __get_user_pages(current, current->mm, addr, 1, | |
1117 | flags, NULL, NULL, NULL); | |
1118 | return rc == -EHWPOISON; | |
1119 | } | |
1120 | ||
af585b92 | 1121 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic, |
612819c3 | 1122 | bool *async, bool write_fault, bool *writable) |
954bbbc2 | 1123 | { |
8d4e1288 | 1124 | struct page *page[1]; |
af585b92 | 1125 | int npages = 0; |
2e2e3738 | 1126 | pfn_t pfn; |
954bbbc2 | 1127 | |
af585b92 GN |
1128 | /* we can do it either atomically or asynchronously, not both */ |
1129 | BUG_ON(atomic && async); | |
1130 | ||
612819c3 MT |
1131 | BUG_ON(!write_fault && !writable); |
1132 | ||
1133 | if (writable) | |
1134 | *writable = true; | |
1135 | ||
af585b92 | 1136 | if (atomic || async) |
887c08ac | 1137 | npages = __get_user_pages_fast(addr, 1, 1, page); |
af585b92 GN |
1138 | |
1139 | if (unlikely(npages != 1) && !atomic) { | |
887c08ac | 1140 | might_sleep(); |
612819c3 MT |
1141 | |
1142 | if (writable) | |
1143 | *writable = write_fault; | |
1144 | ||
0857b9e9 GN |
1145 | if (async) { |
1146 | down_read(¤t->mm->mmap_sem); | |
1147 | npages = get_user_page_nowait(current, current->mm, | |
1148 | addr, write_fault, page); | |
1149 | up_read(¤t->mm->mmap_sem); | |
1150 | } else | |
1151 | npages = get_user_pages_fast(addr, 1, write_fault, | |
1152 | page); | |
612819c3 MT |
1153 | |
1154 | /* map read fault as writable if possible */ | |
1155 | if (unlikely(!write_fault) && npages == 1) { | |
1156 | struct page *wpage[1]; | |
1157 | ||
1158 | npages = __get_user_pages_fast(addr, 1, 1, wpage); | |
1159 | if (npages == 1) { | |
1160 | *writable = true; | |
1161 | put_page(page[0]); | |
1162 | page[0] = wpage[0]; | |
1163 | } | |
1164 | npages = 1; | |
1165 | } | |
887c08ac | 1166 | } |
539cb660 | 1167 | |
2e2e3738 AL |
1168 | if (unlikely(npages != 1)) { |
1169 | struct vm_area_struct *vma; | |
1170 | ||
887c08ac | 1171 | if (atomic) |
8030089f | 1172 | return get_fault_pfn(); |
887c08ac | 1173 | |
bbeb3406 | 1174 | down_read(¤t->mm->mmap_sem); |
0857b9e9 GN |
1175 | if (npages == -EHWPOISON || |
1176 | (!async && check_user_page_hwpoison(addr))) { | |
bbeb3406 | 1177 | up_read(¤t->mm->mmap_sem); |
bf998156 HY |
1178 | get_page(hwpoison_page); |
1179 | return page_to_pfn(hwpoison_page); | |
1180 | } | |
1181 | ||
8030089f | 1182 | vma = find_vma_intersection(current->mm, addr, addr+1); |
4c2155ce | 1183 | |
8030089f GN |
1184 | if (vma == NULL) |
1185 | pfn = get_fault_pfn(); | |
1186 | else if ((vma->vm_flags & VM_PFNMAP)) { | |
1187 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + | |
1188 | vma->vm_pgoff; | |
1189 | BUG_ON(!kvm_is_mmio_pfn(pfn)); | |
1190 | } else { | |
1191 | if (async && (vma->vm_flags & VM_WRITE)) | |
af585b92 | 1192 | *async = true; |
8030089f | 1193 | pfn = get_fault_pfn(); |
2e2e3738 | 1194 | } |
4c2155ce | 1195 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1196 | } else |
1197 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1198 | |
2e2e3738 | 1199 | return pfn; |
35149e21 AL |
1200 | } |
1201 | ||
887c08ac XG |
1202 | pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr) |
1203 | { | |
612819c3 | 1204 | return hva_to_pfn(kvm, addr, true, NULL, true, NULL); |
887c08ac XG |
1205 | } |
1206 | EXPORT_SYMBOL_GPL(hva_to_pfn_atomic); | |
1207 | ||
612819c3 MT |
1208 | static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async, |
1209 | bool write_fault, bool *writable) | |
506f0d6f MT |
1210 | { |
1211 | unsigned long addr; | |
1212 | ||
af585b92 GN |
1213 | if (async) |
1214 | *async = false; | |
1215 | ||
506f0d6f MT |
1216 | addr = gfn_to_hva(kvm, gfn); |
1217 | if (kvm_is_error_hva(addr)) { | |
1218 | get_page(bad_page); | |
1219 | return page_to_pfn(bad_page); | |
1220 | } | |
1221 | ||
612819c3 | 1222 | return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable); |
365fb3fd XG |
1223 | } |
1224 | ||
1225 | pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn) | |
1226 | { | |
612819c3 | 1227 | return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL); |
365fb3fd XG |
1228 | } |
1229 | EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic); | |
1230 | ||
612819c3 MT |
1231 | pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async, |
1232 | bool write_fault, bool *writable) | |
af585b92 | 1233 | { |
612819c3 | 1234 | return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable); |
af585b92 GN |
1235 | } |
1236 | EXPORT_SYMBOL_GPL(gfn_to_pfn_async); | |
1237 | ||
365fb3fd XG |
1238 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
1239 | { | |
612819c3 | 1240 | return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL); |
506f0d6f | 1241 | } |
35149e21 AL |
1242 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
1243 | ||
612819c3 MT |
1244 | pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, |
1245 | bool *writable) | |
1246 | { | |
1247 | return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable); | |
1248 | } | |
1249 | EXPORT_SYMBOL_GPL(gfn_to_pfn_prot); | |
1250 | ||
506f0d6f MT |
1251 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
1252 | struct kvm_memory_slot *slot, gfn_t gfn) | |
1253 | { | |
1254 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
612819c3 | 1255 | return hva_to_pfn(kvm, addr, false, NULL, true, NULL); |
506f0d6f MT |
1256 | } |
1257 | ||
48987781 XG |
1258 | int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages, |
1259 | int nr_pages) | |
1260 | { | |
1261 | unsigned long addr; | |
1262 | gfn_t entry; | |
1263 | ||
49c7754c | 1264 | addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry); |
48987781 XG |
1265 | if (kvm_is_error_hva(addr)) |
1266 | return -1; | |
1267 | ||
1268 | if (entry < nr_pages) | |
1269 | return 0; | |
1270 | ||
1271 | return __get_user_pages_fast(addr, nr_pages, 1, pages); | |
1272 | } | |
1273 | EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic); | |
1274 | ||
35149e21 AL |
1275 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1276 | { | |
2e2e3738 AL |
1277 | pfn_t pfn; |
1278 | ||
1279 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1280 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1281 | return pfn_to_page(pfn); |
1282 | ||
c77fb9dc | 1283 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1284 | |
1285 | get_page(bad_page); | |
1286 | return bad_page; | |
954bbbc2 | 1287 | } |
aab61cc0 | 1288 | |
954bbbc2 AK |
1289 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1290 | ||
b4231d61 IE |
1291 | void kvm_release_page_clean(struct page *page) |
1292 | { | |
35149e21 | 1293 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1294 | } |
1295 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1296 | ||
35149e21 AL |
1297 | void kvm_release_pfn_clean(pfn_t pfn) |
1298 | { | |
c77fb9dc | 1299 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1300 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1301 | } |
1302 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1303 | ||
b4231d61 | 1304 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1305 | { |
35149e21 AL |
1306 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1307 | } | |
1308 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1309 | ||
1310 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1311 | { | |
1312 | kvm_set_pfn_dirty(pfn); | |
1313 | kvm_release_pfn_clean(pfn); | |
1314 | } | |
1315 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1316 | ||
1317 | void kvm_set_page_dirty(struct page *page) | |
1318 | { | |
1319 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1320 | } | |
1321 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1322 | ||
1323 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1324 | { | |
c77fb9dc | 1325 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1326 | struct page *page = pfn_to_page(pfn); |
1327 | if (!PageReserved(page)) | |
1328 | SetPageDirty(page); | |
1329 | } | |
8a7ae055 | 1330 | } |
35149e21 AL |
1331 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1332 | ||
1333 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1334 | { | |
c77fb9dc | 1335 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1336 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1337 | } |
1338 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1339 | ||
1340 | void kvm_get_pfn(pfn_t pfn) | |
1341 | { | |
c77fb9dc | 1342 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1343 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1344 | } |
1345 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1346 | |
195aefde IE |
1347 | static int next_segment(unsigned long len, int offset) |
1348 | { | |
1349 | if (len > PAGE_SIZE - offset) | |
1350 | return PAGE_SIZE - offset; | |
1351 | else | |
1352 | return len; | |
1353 | } | |
1354 | ||
1355 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1356 | int len) | |
1357 | { | |
e0506bcb IE |
1358 | int r; |
1359 | unsigned long addr; | |
195aefde | 1360 | |
e0506bcb IE |
1361 | addr = gfn_to_hva(kvm, gfn); |
1362 | if (kvm_is_error_hva(addr)) | |
1363 | return -EFAULT; | |
fa3d315a | 1364 | r = __copy_from_user(data, (void __user *)addr + offset, len); |
e0506bcb | 1365 | if (r) |
195aefde | 1366 | return -EFAULT; |
195aefde IE |
1367 | return 0; |
1368 | } | |
1369 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1370 | ||
1371 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1372 | { | |
1373 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1374 | int seg; | |
1375 | int offset = offset_in_page(gpa); | |
1376 | int ret; | |
1377 | ||
1378 | while ((seg = next_segment(len, offset)) != 0) { | |
1379 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1380 | if (ret < 0) | |
1381 | return ret; | |
1382 | offset = 0; | |
1383 | len -= seg; | |
1384 | data += seg; | |
1385 | ++gfn; | |
1386 | } | |
1387 | return 0; | |
1388 | } | |
1389 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1390 | ||
7ec54588 MT |
1391 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1392 | unsigned long len) | |
1393 | { | |
1394 | int r; | |
1395 | unsigned long addr; | |
1396 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1397 | int offset = offset_in_page(gpa); | |
1398 | ||
1399 | addr = gfn_to_hva(kvm, gfn); | |
1400 | if (kvm_is_error_hva(addr)) | |
1401 | return -EFAULT; | |
0aac03f0 | 1402 | pagefault_disable(); |
7ec54588 | 1403 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1404 | pagefault_enable(); |
7ec54588 MT |
1405 | if (r) |
1406 | return -EFAULT; | |
1407 | return 0; | |
1408 | } | |
1409 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1410 | ||
195aefde IE |
1411 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1412 | int offset, int len) | |
1413 | { | |
e0506bcb IE |
1414 | int r; |
1415 | unsigned long addr; | |
195aefde | 1416 | |
e0506bcb IE |
1417 | addr = gfn_to_hva(kvm, gfn); |
1418 | if (kvm_is_error_hva(addr)) | |
1419 | return -EFAULT; | |
8b0cedff | 1420 | r = __copy_to_user((void __user *)addr + offset, data, len); |
e0506bcb | 1421 | if (r) |
195aefde | 1422 | return -EFAULT; |
195aefde IE |
1423 | mark_page_dirty(kvm, gfn); |
1424 | return 0; | |
1425 | } | |
1426 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1427 | ||
1428 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1429 | unsigned long len) | |
1430 | { | |
1431 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1432 | int seg; | |
1433 | int offset = offset_in_page(gpa); | |
1434 | int ret; | |
1435 | ||
1436 | while ((seg = next_segment(len, offset)) != 0) { | |
1437 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1438 | if (ret < 0) | |
1439 | return ret; | |
1440 | offset = 0; | |
1441 | len -= seg; | |
1442 | data += seg; | |
1443 | ++gfn; | |
1444 | } | |
1445 | return 0; | |
1446 | } | |
1447 | ||
49c7754c GN |
1448 | int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1449 | gpa_t gpa) | |
1450 | { | |
1451 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1452 | int offset = offset_in_page(gpa); | |
1453 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1454 | ||
1455 | ghc->gpa = gpa; | |
1456 | ghc->generation = slots->generation; | |
1457 | ghc->memslot = __gfn_to_memslot(slots, gfn); | |
1458 | ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL); | |
1459 | if (!kvm_is_error_hva(ghc->hva)) | |
1460 | ghc->hva += offset; | |
1461 | else | |
1462 | return -EFAULT; | |
1463 | ||
1464 | return 0; | |
1465 | } | |
1466 | EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init); | |
1467 | ||
1468 | int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, | |
1469 | void *data, unsigned long len) | |
1470 | { | |
1471 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1472 | int r; | |
1473 | ||
1474 | if (slots->generation != ghc->generation) | |
1475 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1476 | ||
1477 | if (kvm_is_error_hva(ghc->hva)) | |
1478 | return -EFAULT; | |
1479 | ||
8b0cedff | 1480 | r = __copy_to_user((void __user *)ghc->hva, data, len); |
49c7754c GN |
1481 | if (r) |
1482 | return -EFAULT; | |
1483 | mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT); | |
1484 | ||
1485 | return 0; | |
1486 | } | |
1487 | EXPORT_SYMBOL_GPL(kvm_write_guest_cached); | |
1488 | ||
e03b644f GN |
1489 | int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1490 | void *data, unsigned long len) | |
1491 | { | |
1492 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1493 | int r; | |
1494 | ||
1495 | if (slots->generation != ghc->generation) | |
1496 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1497 | ||
1498 | if (kvm_is_error_hva(ghc->hva)) | |
1499 | return -EFAULT; | |
1500 | ||
1501 | r = __copy_from_user(data, (void __user *)ghc->hva, len); | |
1502 | if (r) | |
1503 | return -EFAULT; | |
1504 | ||
1505 | return 0; | |
1506 | } | |
1507 | EXPORT_SYMBOL_GPL(kvm_read_guest_cached); | |
1508 | ||
195aefde IE |
1509 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) |
1510 | { | |
3bcc8a8c HC |
1511 | return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page, |
1512 | offset, len); | |
195aefde IE |
1513 | } |
1514 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1515 | ||
1516 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1517 | { | |
1518 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1519 | int seg; | |
1520 | int offset = offset_in_page(gpa); | |
1521 | int ret; | |
1522 | ||
1523 | while ((seg = next_segment(len, offset)) != 0) { | |
1524 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1525 | if (ret < 0) | |
1526 | return ret; | |
1527 | offset = 0; | |
1528 | len -= seg; | |
1529 | ++gfn; | |
1530 | } | |
1531 | return 0; | |
1532 | } | |
1533 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1534 | ||
49c7754c GN |
1535 | void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, |
1536 | gfn_t gfn) | |
6aa8b732 | 1537 | { |
7e9d619d RR |
1538 | if (memslot && memslot->dirty_bitmap) { |
1539 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1540 | |
7850ac54 TY |
1541 | if (!__test_and_set_bit_le(rel_gfn, memslot->dirty_bitmap)) |
1542 | memslot->nr_dirty_pages++; | |
6aa8b732 AK |
1543 | } |
1544 | } | |
1545 | ||
49c7754c GN |
1546 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1547 | { | |
1548 | struct kvm_memory_slot *memslot; | |
1549 | ||
1550 | memslot = gfn_to_memslot(kvm, gfn); | |
1551 | mark_page_dirty_in_slot(kvm, memslot, gfn); | |
1552 | } | |
1553 | ||
b6958ce4 ED |
1554 | /* |
1555 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1556 | */ | |
8776e519 | 1557 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1558 | { |
e5c239cf MT |
1559 | DEFINE_WAIT(wait); |
1560 | ||
1561 | for (;;) { | |
1562 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1563 | ||
a1b37100 | 1564 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1565 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1566 | break; |
d7690175 | 1567 | } |
09cec754 GN |
1568 | if (kvm_cpu_has_pending_timer(vcpu)) |
1569 | break; | |
e5c239cf MT |
1570 | if (signal_pending(current)) |
1571 | break; | |
1572 | ||
b6958ce4 | 1573 | schedule(); |
b6958ce4 | 1574 | } |
d3bef15f | 1575 | |
e5c239cf | 1576 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1577 | } |
1578 | ||
6aa8b732 AK |
1579 | void kvm_resched(struct kvm_vcpu *vcpu) |
1580 | { | |
3fca0365 YD |
1581 | if (!need_resched()) |
1582 | return; | |
6aa8b732 | 1583 | cond_resched(); |
6aa8b732 AK |
1584 | } |
1585 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1586 | ||
217ece61 | 1587 | void kvm_vcpu_on_spin(struct kvm_vcpu *me) |
d255f4f2 | 1588 | { |
217ece61 RR |
1589 | struct kvm *kvm = me->kvm; |
1590 | struct kvm_vcpu *vcpu; | |
1591 | int last_boosted_vcpu = me->kvm->last_boosted_vcpu; | |
1592 | int yielded = 0; | |
1593 | int pass; | |
1594 | int i; | |
d255f4f2 | 1595 | |
217ece61 RR |
1596 | /* |
1597 | * We boost the priority of a VCPU that is runnable but not | |
1598 | * currently running, because it got preempted by something | |
1599 | * else and called schedule in __vcpu_run. Hopefully that | |
1600 | * VCPU is holding the lock that we need and will release it. | |
1601 | * We approximate round-robin by starting at the last boosted VCPU. | |
1602 | */ | |
1603 | for (pass = 0; pass < 2 && !yielded; pass++) { | |
1604 | kvm_for_each_vcpu(i, vcpu, kvm) { | |
1605 | struct task_struct *task = NULL; | |
1606 | struct pid *pid; | |
1607 | if (!pass && i < last_boosted_vcpu) { | |
1608 | i = last_boosted_vcpu; | |
1609 | continue; | |
1610 | } else if (pass && i > last_boosted_vcpu) | |
1611 | break; | |
1612 | if (vcpu == me) | |
1613 | continue; | |
1614 | if (waitqueue_active(&vcpu->wq)) | |
1615 | continue; | |
1616 | rcu_read_lock(); | |
1617 | pid = rcu_dereference(vcpu->pid); | |
1618 | if (pid) | |
1619 | task = get_pid_task(vcpu->pid, PIDTYPE_PID); | |
1620 | rcu_read_unlock(); | |
1621 | if (!task) | |
1622 | continue; | |
1623 | if (task->flags & PF_VCPU) { | |
1624 | put_task_struct(task); | |
1625 | continue; | |
1626 | } | |
1627 | if (yield_to(task, 1)) { | |
1628 | put_task_struct(task); | |
1629 | kvm->last_boosted_vcpu = i; | |
1630 | yielded = 1; | |
1631 | break; | |
1632 | } | |
1633 | put_task_struct(task); | |
1634 | } | |
1635 | } | |
d255f4f2 ZE |
1636 | } |
1637 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1638 | ||
e4a533a4 | 1639 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1640 | { |
1641 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1642 | struct page *page; |
1643 | ||
e4a533a4 | 1644 | if (vmf->pgoff == 0) |
039576c0 | 1645 | page = virt_to_page(vcpu->run); |
09566765 | 1646 | #ifdef CONFIG_X86 |
e4a533a4 | 1647 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1648 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1649 | #endif |
1650 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1651 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1652 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1653 | #endif |
039576c0 | 1654 | else |
e4a533a4 | 1655 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1656 | get_page(page); |
e4a533a4 NP |
1657 | vmf->page = page; |
1658 | return 0; | |
9a2bb7f4 AK |
1659 | } |
1660 | ||
f0f37e2f | 1661 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1662 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1663 | }; |
1664 | ||
1665 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1666 | { | |
1667 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1668 | return 0; | |
1669 | } | |
1670 | ||
bccf2150 AK |
1671 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1672 | { | |
1673 | struct kvm_vcpu *vcpu = filp->private_data; | |
1674 | ||
66c0b394 | 1675 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1676 | return 0; |
1677 | } | |
1678 | ||
3d3aab1b | 1679 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1680 | .release = kvm_vcpu_release, |
1681 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1dda606c AG |
1682 | #ifdef CONFIG_COMPAT |
1683 | .compat_ioctl = kvm_vcpu_compat_ioctl, | |
1684 | #endif | |
9a2bb7f4 | 1685 | .mmap = kvm_vcpu_mmap, |
6038f373 | 1686 | .llseek = noop_llseek, |
bccf2150 AK |
1687 | }; |
1688 | ||
1689 | /* | |
1690 | * Allocates an inode for the vcpu. | |
1691 | */ | |
1692 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1693 | { | |
628ff7c1 | 1694 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1695 | } |
1696 | ||
c5ea7660 AK |
1697 | /* |
1698 | * Creates some virtual cpus. Good luck creating more than one. | |
1699 | */ | |
73880c80 | 1700 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1701 | { |
1702 | int r; | |
988a2cae | 1703 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1704 | |
73880c80 | 1705 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1706 | if (IS_ERR(vcpu)) |
1707 | return PTR_ERR(vcpu); | |
c5ea7660 | 1708 | |
15ad7146 AK |
1709 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1710 | ||
26e5215f AK |
1711 | r = kvm_arch_vcpu_setup(vcpu); |
1712 | if (r) | |
d780592b | 1713 | goto vcpu_destroy; |
26e5215f | 1714 | |
11ec2804 | 1715 | mutex_lock(&kvm->lock); |
73880c80 GN |
1716 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1717 | r = -EINVAL; | |
d780592b | 1718 | goto unlock_vcpu_destroy; |
fb3f0f51 | 1719 | } |
73880c80 | 1720 | |
988a2cae GN |
1721 | kvm_for_each_vcpu(r, v, kvm) |
1722 | if (v->vcpu_id == id) { | |
73880c80 | 1723 | r = -EEXIST; |
d780592b | 1724 | goto unlock_vcpu_destroy; |
73880c80 GN |
1725 | } |
1726 | ||
1727 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1728 | |
fb3f0f51 | 1729 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1730 | kvm_get_kvm(kvm); |
bccf2150 | 1731 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1732 | if (r < 0) { |
1733 | kvm_put_kvm(kvm); | |
d780592b | 1734 | goto unlock_vcpu_destroy; |
73880c80 GN |
1735 | } |
1736 | ||
1737 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1738 | smp_wmb(); | |
1739 | atomic_inc(&kvm->online_vcpus); | |
1740 | ||
1741 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1742 | if (kvm->bsp_vcpu_id == id) | |
1743 | kvm->bsp_vcpu = vcpu; | |
1744 | #endif | |
1745 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1746 | return r; |
39c3b86e | 1747 | |
d780592b | 1748 | unlock_vcpu_destroy: |
7d8fece6 | 1749 | mutex_unlock(&kvm->lock); |
d780592b | 1750 | vcpu_destroy: |
d40ccc62 | 1751 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1752 | return r; |
1753 | } | |
1754 | ||
1961d276 AK |
1755 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1756 | { | |
1757 | if (sigset) { | |
1758 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1759 | vcpu->sigset_active = 1; | |
1760 | vcpu->sigset = *sigset; | |
1761 | } else | |
1762 | vcpu->sigset_active = 0; | |
1763 | return 0; | |
1764 | } | |
1765 | ||
bccf2150 AK |
1766 | static long kvm_vcpu_ioctl(struct file *filp, |
1767 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1768 | { |
bccf2150 | 1769 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1770 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1771 | int r; |
fa3795a7 DH |
1772 | struct kvm_fpu *fpu = NULL; |
1773 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1774 | |
6d4e4c4f AK |
1775 | if (vcpu->kvm->mm != current->mm) |
1776 | return -EIO; | |
2122ff5e AK |
1777 | |
1778 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1779 | /* | |
1780 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1781 | * so vcpu_load() would break it. | |
1782 | */ | |
1783 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1784 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1785 | #endif | |
1786 | ||
1787 | ||
1788 | vcpu_load(vcpu); | |
6aa8b732 | 1789 | switch (ioctl) { |
9a2bb7f4 | 1790 | case KVM_RUN: |
f0fe5108 AK |
1791 | r = -EINVAL; |
1792 | if (arg) | |
1793 | goto out; | |
b6c7a5dc | 1794 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
64be5007 | 1795 | trace_kvm_userspace_exit(vcpu->run->exit_reason, r); |
6aa8b732 | 1796 | break; |
6aa8b732 | 1797 | case KVM_GET_REGS: { |
3e4bb3ac | 1798 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1799 | |
3e4bb3ac XZ |
1800 | r = -ENOMEM; |
1801 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1802 | if (!kvm_regs) | |
6aa8b732 | 1803 | goto out; |
3e4bb3ac XZ |
1804 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1805 | if (r) | |
1806 | goto out_free1; | |
6aa8b732 | 1807 | r = -EFAULT; |
3e4bb3ac XZ |
1808 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1809 | goto out_free1; | |
6aa8b732 | 1810 | r = 0; |
3e4bb3ac XZ |
1811 | out_free1: |
1812 | kfree(kvm_regs); | |
6aa8b732 AK |
1813 | break; |
1814 | } | |
1815 | case KVM_SET_REGS: { | |
3e4bb3ac | 1816 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1817 | |
3e4bb3ac XZ |
1818 | r = -ENOMEM; |
1819 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1820 | if (!kvm_regs) | |
6aa8b732 | 1821 | goto out; |
3e4bb3ac XZ |
1822 | r = -EFAULT; |
1823 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1824 | goto out_free2; | |
1825 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1826 | if (r) |
3e4bb3ac | 1827 | goto out_free2; |
6aa8b732 | 1828 | r = 0; |
3e4bb3ac XZ |
1829 | out_free2: |
1830 | kfree(kvm_regs); | |
6aa8b732 AK |
1831 | break; |
1832 | } | |
1833 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1834 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1835 | r = -ENOMEM; | |
1836 | if (!kvm_sregs) | |
1837 | goto out; | |
1838 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1839 | if (r) |
1840 | goto out; | |
1841 | r = -EFAULT; | |
fa3795a7 | 1842 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1843 | goto out; |
1844 | r = 0; | |
1845 | break; | |
1846 | } | |
1847 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1848 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1849 | r = -ENOMEM; | |
1850 | if (!kvm_sregs) | |
1851 | goto out; | |
6aa8b732 | 1852 | r = -EFAULT; |
fa3795a7 | 1853 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1854 | goto out; |
fa3795a7 | 1855 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1856 | if (r) |
1857 | goto out; | |
1858 | r = 0; | |
1859 | break; | |
1860 | } | |
62d9f0db MT |
1861 | case KVM_GET_MP_STATE: { |
1862 | struct kvm_mp_state mp_state; | |
1863 | ||
1864 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1865 | if (r) | |
1866 | goto out; | |
1867 | r = -EFAULT; | |
1868 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1869 | goto out; | |
1870 | r = 0; | |
1871 | break; | |
1872 | } | |
1873 | case KVM_SET_MP_STATE: { | |
1874 | struct kvm_mp_state mp_state; | |
1875 | ||
1876 | r = -EFAULT; | |
1877 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1878 | goto out; | |
1879 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1880 | if (r) | |
1881 | goto out; | |
1882 | r = 0; | |
1883 | break; | |
1884 | } | |
6aa8b732 AK |
1885 | case KVM_TRANSLATE: { |
1886 | struct kvm_translation tr; | |
1887 | ||
1888 | r = -EFAULT; | |
2f366987 | 1889 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1890 | goto out; |
8b006791 | 1891 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1892 | if (r) |
1893 | goto out; | |
1894 | r = -EFAULT; | |
2f366987 | 1895 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1896 | goto out; |
1897 | r = 0; | |
1898 | break; | |
1899 | } | |
d0bfb940 JK |
1900 | case KVM_SET_GUEST_DEBUG: { |
1901 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1902 | |
1903 | r = -EFAULT; | |
2f366987 | 1904 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1905 | goto out; |
d0bfb940 | 1906 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1907 | if (r) |
1908 | goto out; | |
1909 | r = 0; | |
1910 | break; | |
1911 | } | |
1961d276 AK |
1912 | case KVM_SET_SIGNAL_MASK: { |
1913 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1914 | struct kvm_signal_mask kvm_sigmask; | |
1915 | sigset_t sigset, *p; | |
1916 | ||
1917 | p = NULL; | |
1918 | if (argp) { | |
1919 | r = -EFAULT; | |
1920 | if (copy_from_user(&kvm_sigmask, argp, | |
1921 | sizeof kvm_sigmask)) | |
1922 | goto out; | |
1923 | r = -EINVAL; | |
1924 | if (kvm_sigmask.len != sizeof sigset) | |
1925 | goto out; | |
1926 | r = -EFAULT; | |
1927 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1928 | sizeof sigset)) | |
1929 | goto out; | |
1930 | p = &sigset; | |
1931 | } | |
376d41ff | 1932 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1933 | break; |
1934 | } | |
b8836737 | 1935 | case KVM_GET_FPU: { |
fa3795a7 DH |
1936 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1937 | r = -ENOMEM; | |
1938 | if (!fpu) | |
1939 | goto out; | |
1940 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1941 | if (r) |
1942 | goto out; | |
1943 | r = -EFAULT; | |
fa3795a7 | 1944 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1945 | goto out; |
1946 | r = 0; | |
1947 | break; | |
1948 | } | |
1949 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1950 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1951 | r = -ENOMEM; | |
1952 | if (!fpu) | |
1953 | goto out; | |
b8836737 | 1954 | r = -EFAULT; |
fa3795a7 | 1955 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1956 | goto out; |
fa3795a7 | 1957 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1958 | if (r) |
1959 | goto out; | |
1960 | r = 0; | |
1961 | break; | |
1962 | } | |
bccf2150 | 1963 | default: |
313a3dc7 | 1964 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1965 | } |
1966 | out: | |
2122ff5e | 1967 | vcpu_put(vcpu); |
fa3795a7 DH |
1968 | kfree(fpu); |
1969 | kfree(kvm_sregs); | |
bccf2150 AK |
1970 | return r; |
1971 | } | |
1972 | ||
1dda606c AG |
1973 | #ifdef CONFIG_COMPAT |
1974 | static long kvm_vcpu_compat_ioctl(struct file *filp, | |
1975 | unsigned int ioctl, unsigned long arg) | |
1976 | { | |
1977 | struct kvm_vcpu *vcpu = filp->private_data; | |
1978 | void __user *argp = compat_ptr(arg); | |
1979 | int r; | |
1980 | ||
1981 | if (vcpu->kvm->mm != current->mm) | |
1982 | return -EIO; | |
1983 | ||
1984 | switch (ioctl) { | |
1985 | case KVM_SET_SIGNAL_MASK: { | |
1986 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1987 | struct kvm_signal_mask kvm_sigmask; | |
1988 | compat_sigset_t csigset; | |
1989 | sigset_t sigset; | |
1990 | ||
1991 | if (argp) { | |
1992 | r = -EFAULT; | |
1993 | if (copy_from_user(&kvm_sigmask, argp, | |
1994 | sizeof kvm_sigmask)) | |
1995 | goto out; | |
1996 | r = -EINVAL; | |
1997 | if (kvm_sigmask.len != sizeof csigset) | |
1998 | goto out; | |
1999 | r = -EFAULT; | |
2000 | if (copy_from_user(&csigset, sigmask_arg->sigset, | |
2001 | sizeof csigset)) | |
2002 | goto out; | |
2003 | } | |
2004 | sigset_from_compat(&sigset, &csigset); | |
2005 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
2006 | break; | |
2007 | } | |
2008 | default: | |
2009 | r = kvm_vcpu_ioctl(filp, ioctl, arg); | |
2010 | } | |
2011 | ||
2012 | out: | |
2013 | return r; | |
2014 | } | |
2015 | #endif | |
2016 | ||
bccf2150 AK |
2017 | static long kvm_vm_ioctl(struct file *filp, |
2018 | unsigned int ioctl, unsigned long arg) | |
2019 | { | |
2020 | struct kvm *kvm = filp->private_data; | |
2021 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 2022 | int r; |
bccf2150 | 2023 | |
6d4e4c4f AK |
2024 | if (kvm->mm != current->mm) |
2025 | return -EIO; | |
bccf2150 AK |
2026 | switch (ioctl) { |
2027 | case KVM_CREATE_VCPU: | |
2028 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
2029 | if (r < 0) | |
2030 | goto out; | |
2031 | break; | |
6fc138d2 IE |
2032 | case KVM_SET_USER_MEMORY_REGION: { |
2033 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2034 | ||
2035 | r = -EFAULT; | |
2036 | if (copy_from_user(&kvm_userspace_mem, argp, | |
2037 | sizeof kvm_userspace_mem)) | |
2038 | goto out; | |
2039 | ||
2040 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
2041 | if (r) |
2042 | goto out; | |
2043 | break; | |
2044 | } | |
2045 | case KVM_GET_DIRTY_LOG: { | |
2046 | struct kvm_dirty_log log; | |
2047 | ||
2048 | r = -EFAULT; | |
2f366987 | 2049 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2050 | goto out; |
2c6f5df9 | 2051 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2052 | if (r) |
2053 | goto out; | |
2054 | break; | |
2055 | } | |
5f94c174 LV |
2056 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2057 | case KVM_REGISTER_COALESCED_MMIO: { | |
2058 | struct kvm_coalesced_mmio_zone zone; | |
2059 | r = -EFAULT; | |
2060 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2061 | goto out; | |
5f94c174 LV |
2062 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
2063 | if (r) | |
2064 | goto out; | |
2065 | r = 0; | |
2066 | break; | |
2067 | } | |
2068 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
2069 | struct kvm_coalesced_mmio_zone zone; | |
2070 | r = -EFAULT; | |
2071 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2072 | goto out; | |
5f94c174 LV |
2073 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
2074 | if (r) | |
2075 | goto out; | |
2076 | r = 0; | |
2077 | break; | |
2078 | } | |
2079 | #endif | |
721eecbf GH |
2080 | case KVM_IRQFD: { |
2081 | struct kvm_irqfd data; | |
2082 | ||
2083 | r = -EFAULT; | |
2084 | if (copy_from_user(&data, argp, sizeof data)) | |
2085 | goto out; | |
2086 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
2087 | break; | |
2088 | } | |
d34e6b17 GH |
2089 | case KVM_IOEVENTFD: { |
2090 | struct kvm_ioeventfd data; | |
2091 | ||
2092 | r = -EFAULT; | |
2093 | if (copy_from_user(&data, argp, sizeof data)) | |
2094 | goto out; | |
2095 | r = kvm_ioeventfd(kvm, &data); | |
2096 | break; | |
2097 | } | |
73880c80 GN |
2098 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2099 | case KVM_SET_BOOT_CPU_ID: | |
2100 | r = 0; | |
894a9c55 | 2101 | mutex_lock(&kvm->lock); |
73880c80 GN |
2102 | if (atomic_read(&kvm->online_vcpus) != 0) |
2103 | r = -EBUSY; | |
2104 | else | |
2105 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 2106 | mutex_unlock(&kvm->lock); |
73880c80 GN |
2107 | break; |
2108 | #endif | |
f17abe9a | 2109 | default: |
1fe779f8 | 2110 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
2111 | if (r == -ENOTTY) |
2112 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
2113 | } |
2114 | out: | |
2115 | return r; | |
2116 | } | |
2117 | ||
6ff5894c AB |
2118 | #ifdef CONFIG_COMPAT |
2119 | struct compat_kvm_dirty_log { | |
2120 | __u32 slot; | |
2121 | __u32 padding1; | |
2122 | union { | |
2123 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
2124 | __u64 padding2; | |
2125 | }; | |
2126 | }; | |
2127 | ||
2128 | static long kvm_vm_compat_ioctl(struct file *filp, | |
2129 | unsigned int ioctl, unsigned long arg) | |
2130 | { | |
2131 | struct kvm *kvm = filp->private_data; | |
2132 | int r; | |
2133 | ||
2134 | if (kvm->mm != current->mm) | |
2135 | return -EIO; | |
2136 | switch (ioctl) { | |
2137 | case KVM_GET_DIRTY_LOG: { | |
2138 | struct compat_kvm_dirty_log compat_log; | |
2139 | struct kvm_dirty_log log; | |
2140 | ||
2141 | r = -EFAULT; | |
2142 | if (copy_from_user(&compat_log, (void __user *)arg, | |
2143 | sizeof(compat_log))) | |
2144 | goto out; | |
2145 | log.slot = compat_log.slot; | |
2146 | log.padding1 = compat_log.padding1; | |
2147 | log.padding2 = compat_log.padding2; | |
2148 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
2149 | ||
2150 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
2151 | if (r) | |
2152 | goto out; | |
2153 | break; | |
2154 | } | |
2155 | default: | |
2156 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
2157 | } | |
2158 | ||
2159 | out: | |
2160 | return r; | |
2161 | } | |
2162 | #endif | |
2163 | ||
e4a533a4 | 2164 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 2165 | { |
777b3f49 MT |
2166 | struct page *page[1]; |
2167 | unsigned long addr; | |
2168 | int npages; | |
2169 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 2170 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 2171 | |
777b3f49 MT |
2172 | addr = gfn_to_hva(kvm, gfn); |
2173 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 2174 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2175 | |
2176 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
2177 | NULL); | |
2178 | if (unlikely(npages != 1)) | |
e4a533a4 | 2179 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2180 | |
2181 | vmf->page = page[0]; | |
e4a533a4 | 2182 | return 0; |
f17abe9a AK |
2183 | } |
2184 | ||
f0f37e2f | 2185 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 2186 | .fault = kvm_vm_fault, |
f17abe9a AK |
2187 | }; |
2188 | ||
2189 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2190 | { | |
2191 | vma->vm_ops = &kvm_vm_vm_ops; | |
2192 | return 0; | |
2193 | } | |
2194 | ||
3d3aab1b | 2195 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
2196 | .release = kvm_vm_release, |
2197 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
2198 | #ifdef CONFIG_COMPAT |
2199 | .compat_ioctl = kvm_vm_compat_ioctl, | |
2200 | #endif | |
f17abe9a | 2201 | .mmap = kvm_vm_mmap, |
6038f373 | 2202 | .llseek = noop_llseek, |
f17abe9a AK |
2203 | }; |
2204 | ||
2205 | static int kvm_dev_ioctl_create_vm(void) | |
2206 | { | |
aac87636 | 2207 | int r; |
f17abe9a AK |
2208 | struct kvm *kvm; |
2209 | ||
f17abe9a | 2210 | kvm = kvm_create_vm(); |
d6d28168 AK |
2211 | if (IS_ERR(kvm)) |
2212 | return PTR_ERR(kvm); | |
6ce5a090 TY |
2213 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2214 | r = kvm_coalesced_mmio_init(kvm); | |
2215 | if (r < 0) { | |
2216 | kvm_put_kvm(kvm); | |
2217 | return r; | |
2218 | } | |
2219 | #endif | |
aac87636 HC |
2220 | r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2221 | if (r < 0) | |
66c0b394 | 2222 | kvm_put_kvm(kvm); |
f17abe9a | 2223 | |
aac87636 | 2224 | return r; |
f17abe9a AK |
2225 | } |
2226 | ||
1a811b61 AK |
2227 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2228 | { | |
2229 | switch (arg) { | |
ca9edaee | 2230 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2231 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2232 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
2233 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2234 | case KVM_CAP_SET_BOOT_CPU_ID: | |
2235 | #endif | |
a9c7399d | 2236 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 2237 | return 1; |
399ec807 AK |
2238 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
2239 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 2240 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2241 | #endif |
1a811b61 AK |
2242 | default: |
2243 | break; | |
2244 | } | |
2245 | return kvm_dev_ioctl_check_extension(arg); | |
2246 | } | |
2247 | ||
f17abe9a AK |
2248 | static long kvm_dev_ioctl(struct file *filp, |
2249 | unsigned int ioctl, unsigned long arg) | |
2250 | { | |
07c45a36 | 2251 | long r = -EINVAL; |
f17abe9a AK |
2252 | |
2253 | switch (ioctl) { | |
2254 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2255 | r = -EINVAL; |
2256 | if (arg) | |
2257 | goto out; | |
f17abe9a AK |
2258 | r = KVM_API_VERSION; |
2259 | break; | |
2260 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2261 | r = -EINVAL; |
2262 | if (arg) | |
2263 | goto out; | |
f17abe9a AK |
2264 | r = kvm_dev_ioctl_create_vm(); |
2265 | break; | |
018d00d2 | 2266 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2267 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2268 | break; |
07c45a36 AK |
2269 | case KVM_GET_VCPU_MMAP_SIZE: |
2270 | r = -EINVAL; | |
2271 | if (arg) | |
2272 | goto out; | |
adb1ff46 AK |
2273 | r = PAGE_SIZE; /* struct kvm_run */ |
2274 | #ifdef CONFIG_X86 | |
2275 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2276 | #endif |
2277 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2278 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2279 | #endif |
07c45a36 | 2280 | break; |
d4c9ff2d FEL |
2281 | case KVM_TRACE_ENABLE: |
2282 | case KVM_TRACE_PAUSE: | |
2283 | case KVM_TRACE_DISABLE: | |
2023a29c | 2284 | r = -EOPNOTSUPP; |
d4c9ff2d | 2285 | break; |
6aa8b732 | 2286 | default: |
043405e1 | 2287 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2288 | } |
2289 | out: | |
2290 | return r; | |
2291 | } | |
2292 | ||
6aa8b732 | 2293 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2294 | .unlocked_ioctl = kvm_dev_ioctl, |
2295 | .compat_ioctl = kvm_dev_ioctl, | |
6038f373 | 2296 | .llseek = noop_llseek, |
6aa8b732 AK |
2297 | }; |
2298 | ||
2299 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2300 | KVM_MINOR, |
6aa8b732 AK |
2301 | "kvm", |
2302 | &kvm_chardev_ops, | |
2303 | }; | |
2304 | ||
75b7127c | 2305 | static void hardware_enable_nolock(void *junk) |
1b6c0168 AK |
2306 | { |
2307 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 2308 | int r; |
1b6c0168 | 2309 | |
7f59f492 | 2310 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2311 | return; |
10474ae8 | 2312 | |
7f59f492 | 2313 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
2314 | |
2315 | r = kvm_arch_hardware_enable(NULL); | |
2316 | ||
2317 | if (r) { | |
2318 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
2319 | atomic_inc(&hardware_enable_failed); | |
2320 | printk(KERN_INFO "kvm: enabling virtualization on " | |
2321 | "CPU%d failed\n", cpu); | |
2322 | } | |
1b6c0168 AK |
2323 | } |
2324 | ||
75b7127c TY |
2325 | static void hardware_enable(void *junk) |
2326 | { | |
e935b837 | 2327 | raw_spin_lock(&kvm_lock); |
75b7127c | 2328 | hardware_enable_nolock(junk); |
e935b837 | 2329 | raw_spin_unlock(&kvm_lock); |
75b7127c TY |
2330 | } |
2331 | ||
2332 | static void hardware_disable_nolock(void *junk) | |
1b6c0168 AK |
2333 | { |
2334 | int cpu = raw_smp_processor_id(); | |
2335 | ||
7f59f492 | 2336 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2337 | return; |
7f59f492 | 2338 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2339 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2340 | } |
2341 | ||
75b7127c TY |
2342 | static void hardware_disable(void *junk) |
2343 | { | |
e935b837 | 2344 | raw_spin_lock(&kvm_lock); |
75b7127c | 2345 | hardware_disable_nolock(junk); |
e935b837 | 2346 | raw_spin_unlock(&kvm_lock); |
75b7127c TY |
2347 | } |
2348 | ||
10474ae8 AG |
2349 | static void hardware_disable_all_nolock(void) |
2350 | { | |
2351 | BUG_ON(!kvm_usage_count); | |
2352 | ||
2353 | kvm_usage_count--; | |
2354 | if (!kvm_usage_count) | |
75b7127c | 2355 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
10474ae8 AG |
2356 | } |
2357 | ||
2358 | static void hardware_disable_all(void) | |
2359 | { | |
e935b837 | 2360 | raw_spin_lock(&kvm_lock); |
10474ae8 | 2361 | hardware_disable_all_nolock(); |
e935b837 | 2362 | raw_spin_unlock(&kvm_lock); |
10474ae8 AG |
2363 | } |
2364 | ||
2365 | static int hardware_enable_all(void) | |
2366 | { | |
2367 | int r = 0; | |
2368 | ||
e935b837 | 2369 | raw_spin_lock(&kvm_lock); |
10474ae8 AG |
2370 | |
2371 | kvm_usage_count++; | |
2372 | if (kvm_usage_count == 1) { | |
2373 | atomic_set(&hardware_enable_failed, 0); | |
75b7127c | 2374 | on_each_cpu(hardware_enable_nolock, NULL, 1); |
10474ae8 AG |
2375 | |
2376 | if (atomic_read(&hardware_enable_failed)) { | |
2377 | hardware_disable_all_nolock(); | |
2378 | r = -EBUSY; | |
2379 | } | |
2380 | } | |
2381 | ||
e935b837 | 2382 | raw_spin_unlock(&kvm_lock); |
10474ae8 AG |
2383 | |
2384 | return r; | |
2385 | } | |
2386 | ||
774c47f1 AK |
2387 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2388 | void *v) | |
2389 | { | |
2390 | int cpu = (long)v; | |
2391 | ||
10474ae8 AG |
2392 | if (!kvm_usage_count) |
2393 | return NOTIFY_OK; | |
2394 | ||
1a6f4d7f | 2395 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2396 | switch (val) { |
cec9ad27 | 2397 | case CPU_DYING: |
6ec8a856 AK |
2398 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2399 | cpu); | |
2400 | hardware_disable(NULL); | |
2401 | break; | |
da908f2f | 2402 | case CPU_STARTING: |
43934a38 JK |
2403 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2404 | cpu); | |
da908f2f | 2405 | hardware_enable(NULL); |
774c47f1 AK |
2406 | break; |
2407 | } | |
2408 | return NOTIFY_OK; | |
2409 | } | |
2410 | ||
4ecac3fd | 2411 | |
b7c4145b | 2412 | asmlinkage void kvm_spurious_fault(void) |
4ecac3fd | 2413 | { |
4ecac3fd AK |
2414 | /* Fault while not rebooting. We want the trace. */ |
2415 | BUG(); | |
2416 | } | |
b7c4145b | 2417 | EXPORT_SYMBOL_GPL(kvm_spurious_fault); |
4ecac3fd | 2418 | |
9a2b85c6 | 2419 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2420 | void *v) |
9a2b85c6 | 2421 | { |
8e1c1815 SY |
2422 | /* |
2423 | * Some (well, at least mine) BIOSes hang on reboot if | |
2424 | * in vmx root mode. | |
2425 | * | |
2426 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2427 | */ | |
2428 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2429 | kvm_rebooting = true; | |
75b7127c | 2430 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
9a2b85c6 RR |
2431 | return NOTIFY_OK; |
2432 | } | |
2433 | ||
2434 | static struct notifier_block kvm_reboot_notifier = { | |
2435 | .notifier_call = kvm_reboot, | |
2436 | .priority = 0, | |
2437 | }; | |
2438 | ||
e93f8a0f | 2439 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
2440 | { |
2441 | int i; | |
2442 | ||
2443 | for (i = 0; i < bus->dev_count; i++) { | |
743eeb0b | 2444 | struct kvm_io_device *pos = bus->range[i].dev; |
2eeb2e94 GH |
2445 | |
2446 | kvm_iodevice_destructor(pos); | |
2447 | } | |
e93f8a0f | 2448 | kfree(bus); |
2eeb2e94 GH |
2449 | } |
2450 | ||
743eeb0b SL |
2451 | int kvm_io_bus_sort_cmp(const void *p1, const void *p2) |
2452 | { | |
2453 | const struct kvm_io_range *r1 = p1; | |
2454 | const struct kvm_io_range *r2 = p2; | |
2455 | ||
2456 | if (r1->addr < r2->addr) | |
2457 | return -1; | |
2458 | if (r1->addr + r1->len > r2->addr + r2->len) | |
2459 | return 1; | |
2460 | return 0; | |
2461 | } | |
2462 | ||
2463 | int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev, | |
2464 | gpa_t addr, int len) | |
2465 | { | |
2466 | if (bus->dev_count == NR_IOBUS_DEVS) | |
2467 | return -ENOSPC; | |
2468 | ||
2469 | bus->range[bus->dev_count++] = (struct kvm_io_range) { | |
2470 | .addr = addr, | |
2471 | .len = len, | |
2472 | .dev = dev, | |
2473 | }; | |
2474 | ||
2475 | sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range), | |
2476 | kvm_io_bus_sort_cmp, NULL); | |
2477 | ||
2478 | return 0; | |
2479 | } | |
2480 | ||
2481 | int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus, | |
2482 | gpa_t addr, int len) | |
2483 | { | |
2484 | struct kvm_io_range *range, key; | |
2485 | int off; | |
2486 | ||
2487 | key = (struct kvm_io_range) { | |
2488 | .addr = addr, | |
2489 | .len = len, | |
2490 | }; | |
2491 | ||
2492 | range = bsearch(&key, bus->range, bus->dev_count, | |
2493 | sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp); | |
2494 | if (range == NULL) | |
2495 | return -ENOENT; | |
2496 | ||
2497 | off = range - bus->range; | |
2498 | ||
2499 | while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0) | |
2500 | off--; | |
2501 | ||
2502 | return off; | |
2503 | } | |
2504 | ||
bda9020e | 2505 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2506 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2507 | int len, const void *val) |
2eeb2e94 | 2508 | { |
743eeb0b | 2509 | int idx; |
90d83dc3 | 2510 | struct kvm_io_bus *bus; |
743eeb0b SL |
2511 | struct kvm_io_range range; |
2512 | ||
2513 | range = (struct kvm_io_range) { | |
2514 | .addr = addr, | |
2515 | .len = len, | |
2516 | }; | |
90d83dc3 LJ |
2517 | |
2518 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
743eeb0b SL |
2519 | idx = kvm_io_bus_get_first_dev(bus, addr, len); |
2520 | if (idx < 0) | |
2521 | return -EOPNOTSUPP; | |
2522 | ||
2523 | while (idx < bus->dev_count && | |
2524 | kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) { | |
2525 | if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val)) | |
bda9020e | 2526 | return 0; |
743eeb0b SL |
2527 | idx++; |
2528 | } | |
2529 | ||
bda9020e MT |
2530 | return -EOPNOTSUPP; |
2531 | } | |
2eeb2e94 | 2532 | |
bda9020e | 2533 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2534 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2535 | int len, void *val) | |
bda9020e | 2536 | { |
743eeb0b | 2537 | int idx; |
90d83dc3 | 2538 | struct kvm_io_bus *bus; |
743eeb0b SL |
2539 | struct kvm_io_range range; |
2540 | ||
2541 | range = (struct kvm_io_range) { | |
2542 | .addr = addr, | |
2543 | .len = len, | |
2544 | }; | |
e93f8a0f | 2545 | |
90d83dc3 | 2546 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
743eeb0b SL |
2547 | idx = kvm_io_bus_get_first_dev(bus, addr, len); |
2548 | if (idx < 0) | |
2549 | return -EOPNOTSUPP; | |
2550 | ||
2551 | while (idx < bus->dev_count && | |
2552 | kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) { | |
2553 | if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val)) | |
bda9020e | 2554 | return 0; |
743eeb0b SL |
2555 | idx++; |
2556 | } | |
2557 | ||
bda9020e | 2558 | return -EOPNOTSUPP; |
2eeb2e94 GH |
2559 | } |
2560 | ||
79fac95e | 2561 | /* Caller must hold slots_lock. */ |
743eeb0b SL |
2562 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2563 | int len, struct kvm_io_device *dev) | |
6c474694 | 2564 | { |
e93f8a0f | 2565 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2566 | |
e93f8a0f | 2567 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2568 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2569 | return -ENOSPC; | |
2eeb2e94 | 2570 | |
6da64fdb | 2571 | new_bus = kmemdup(bus, sizeof(struct kvm_io_bus), GFP_KERNEL); |
e93f8a0f MT |
2572 | if (!new_bus) |
2573 | return -ENOMEM; | |
743eeb0b | 2574 | kvm_io_bus_insert_dev(new_bus, dev, addr, len); |
e93f8a0f MT |
2575 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); |
2576 | synchronize_srcu_expedited(&kvm->srcu); | |
2577 | kfree(bus); | |
090b7aff GH |
2578 | |
2579 | return 0; | |
2580 | } | |
2581 | ||
79fac95e | 2582 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2583 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2584 | struct kvm_io_device *dev) | |
090b7aff | 2585 | { |
e93f8a0f MT |
2586 | int i, r; |
2587 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2588 | |
e93f8a0f MT |
2589 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2590 | if (!new_bus) | |
2591 | return -ENOMEM; | |
090b7aff | 2592 | |
e93f8a0f MT |
2593 | bus = kvm->buses[bus_idx]; |
2594 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2595 | ||
2596 | r = -ENOENT; | |
2597 | for (i = 0; i < new_bus->dev_count; i++) | |
743eeb0b | 2598 | if (new_bus->range[i].dev == dev) { |
e93f8a0f | 2599 | r = 0; |
743eeb0b SL |
2600 | new_bus->dev_count--; |
2601 | new_bus->range[i] = new_bus->range[new_bus->dev_count]; | |
2602 | sort(new_bus->range, new_bus->dev_count, | |
2603 | sizeof(struct kvm_io_range), | |
2604 | kvm_io_bus_sort_cmp, NULL); | |
090b7aff GH |
2605 | break; |
2606 | } | |
e93f8a0f MT |
2607 | |
2608 | if (r) { | |
2609 | kfree(new_bus); | |
2610 | return r; | |
2611 | } | |
2612 | ||
2613 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2614 | synchronize_srcu_expedited(&kvm->srcu); | |
2615 | kfree(bus); | |
2616 | return r; | |
2eeb2e94 GH |
2617 | } |
2618 | ||
774c47f1 AK |
2619 | static struct notifier_block kvm_cpu_notifier = { |
2620 | .notifier_call = kvm_cpu_hotplug, | |
774c47f1 AK |
2621 | }; |
2622 | ||
8b88b099 | 2623 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2624 | { |
2625 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2626 | struct kvm *kvm; |
2627 | ||
8b88b099 | 2628 | *val = 0; |
e935b837 | 2629 | raw_spin_lock(&kvm_lock); |
ba1389b7 | 2630 | list_for_each_entry(kvm, &vm_list, vm_list) |
8b88b099 | 2631 | *val += *(u32 *)((void *)kvm + offset); |
e935b837 | 2632 | raw_spin_unlock(&kvm_lock); |
8b88b099 | 2633 | return 0; |
ba1389b7 AK |
2634 | } |
2635 | ||
2636 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2637 | ||
8b88b099 | 2638 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2639 | { |
2640 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2641 | struct kvm *kvm; |
2642 | struct kvm_vcpu *vcpu; | |
2643 | int i; | |
2644 | ||
8b88b099 | 2645 | *val = 0; |
e935b837 | 2646 | raw_spin_lock(&kvm_lock); |
1165f5fe | 2647 | list_for_each_entry(kvm, &vm_list, vm_list) |
988a2cae GN |
2648 | kvm_for_each_vcpu(i, vcpu, kvm) |
2649 | *val += *(u32 *)((void *)vcpu + offset); | |
2650 | ||
e935b837 | 2651 | raw_spin_unlock(&kvm_lock); |
8b88b099 | 2652 | return 0; |
1165f5fe AK |
2653 | } |
2654 | ||
ba1389b7 AK |
2655 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2656 | ||
828c0950 | 2657 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2658 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2659 | [KVM_STAT_VM] = &vm_stat_fops, | |
2660 | }; | |
1165f5fe | 2661 | |
a16b043c | 2662 | static void kvm_init_debug(void) |
6aa8b732 AK |
2663 | { |
2664 | struct kvm_stats_debugfs_item *p; | |
2665 | ||
76f7c879 | 2666 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2667 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2668 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2669 | (void *)(long)p->offset, |
ba1389b7 | 2670 | stat_fops[p->kind]); |
6aa8b732 AK |
2671 | } |
2672 | ||
2673 | static void kvm_exit_debug(void) | |
2674 | { | |
2675 | struct kvm_stats_debugfs_item *p; | |
2676 | ||
2677 | for (p = debugfs_entries; p->name; ++p) | |
2678 | debugfs_remove(p->dentry); | |
76f7c879 | 2679 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2680 | } |
2681 | ||
fb3600cc | 2682 | static int kvm_suspend(void) |
59ae6c6b | 2683 | { |
10474ae8 | 2684 | if (kvm_usage_count) |
75b7127c | 2685 | hardware_disable_nolock(NULL); |
59ae6c6b AK |
2686 | return 0; |
2687 | } | |
2688 | ||
fb3600cc | 2689 | static void kvm_resume(void) |
59ae6c6b | 2690 | { |
ca84d1a2 | 2691 | if (kvm_usage_count) { |
e935b837 | 2692 | WARN_ON(raw_spin_is_locked(&kvm_lock)); |
75b7127c | 2693 | hardware_enable_nolock(NULL); |
ca84d1a2 | 2694 | } |
59ae6c6b AK |
2695 | } |
2696 | ||
fb3600cc | 2697 | static struct syscore_ops kvm_syscore_ops = { |
59ae6c6b AK |
2698 | .suspend = kvm_suspend, |
2699 | .resume = kvm_resume, | |
2700 | }; | |
2701 | ||
cea7bb21 | 2702 | struct page *bad_page; |
35149e21 | 2703 | pfn_t bad_pfn; |
6aa8b732 | 2704 | |
15ad7146 AK |
2705 | static inline |
2706 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2707 | { | |
2708 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2709 | } | |
2710 | ||
2711 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2712 | { | |
2713 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2714 | ||
e9b11c17 | 2715 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2716 | } |
2717 | ||
2718 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2719 | struct task_struct *next) | |
2720 | { | |
2721 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2722 | ||
e9b11c17 | 2723 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2724 | } |
2725 | ||
0ee75bea | 2726 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2727 | struct module *module) |
6aa8b732 AK |
2728 | { |
2729 | int r; | |
002c7f7c | 2730 | int cpu; |
6aa8b732 | 2731 | |
f8c16bba ZX |
2732 | r = kvm_arch_init(opaque); |
2733 | if (r) | |
d2308784 | 2734 | goto out_fail; |
cb498ea2 ZX |
2735 | |
2736 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2737 | ||
2738 | if (bad_page == NULL) { | |
2739 | r = -ENOMEM; | |
2740 | goto out; | |
2741 | } | |
2742 | ||
35149e21 AL |
2743 | bad_pfn = page_to_pfn(bad_page); |
2744 | ||
bf998156 HY |
2745 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2746 | ||
2747 | if (hwpoison_page == NULL) { | |
2748 | r = -ENOMEM; | |
2749 | goto out_free_0; | |
2750 | } | |
2751 | ||
2752 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2753 | ||
edba23e5 GN |
2754 | fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2755 | ||
2756 | if (fault_page == NULL) { | |
2757 | r = -ENOMEM; | |
2758 | goto out_free_0; | |
2759 | } | |
2760 | ||
2761 | fault_pfn = page_to_pfn(fault_page); | |
2762 | ||
8437a617 | 2763 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2764 | r = -ENOMEM; |
2765 | goto out_free_0; | |
2766 | } | |
2767 | ||
e9b11c17 | 2768 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2769 | if (r < 0) |
7f59f492 | 2770 | goto out_free_0a; |
6aa8b732 | 2771 | |
002c7f7c YS |
2772 | for_each_online_cpu(cpu) { |
2773 | smp_call_function_single(cpu, | |
e9b11c17 | 2774 | kvm_arch_check_processor_compat, |
8691e5a8 | 2775 | &r, 1); |
002c7f7c | 2776 | if (r < 0) |
d2308784 | 2777 | goto out_free_1; |
002c7f7c YS |
2778 | } |
2779 | ||
774c47f1 AK |
2780 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2781 | if (r) | |
d2308784 | 2782 | goto out_free_2; |
6aa8b732 AK |
2783 | register_reboot_notifier(&kvm_reboot_notifier); |
2784 | ||
c16f862d | 2785 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2786 | if (!vcpu_align) |
2787 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2788 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2789 | 0, NULL); |
c16f862d RR |
2790 | if (!kvm_vcpu_cache) { |
2791 | r = -ENOMEM; | |
fb3600cc | 2792 | goto out_free_3; |
c16f862d RR |
2793 | } |
2794 | ||
af585b92 GN |
2795 | r = kvm_async_pf_init(); |
2796 | if (r) | |
2797 | goto out_free; | |
2798 | ||
6aa8b732 | 2799 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2800 | kvm_vm_fops.owner = module; |
2801 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2802 | |
2803 | r = misc_register(&kvm_dev); | |
2804 | if (r) { | |
d77c26fc | 2805 | printk(KERN_ERR "kvm: misc device register failed\n"); |
af585b92 | 2806 | goto out_unreg; |
6aa8b732 AK |
2807 | } |
2808 | ||
fb3600cc RW |
2809 | register_syscore_ops(&kvm_syscore_ops); |
2810 | ||
15ad7146 AK |
2811 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2812 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2813 | ||
0ea4ed8e DW |
2814 | kvm_init_debug(); |
2815 | ||
c7addb90 | 2816 | return 0; |
6aa8b732 | 2817 | |
af585b92 GN |
2818 | out_unreg: |
2819 | kvm_async_pf_deinit(); | |
6aa8b732 | 2820 | out_free: |
c16f862d | 2821 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2822 | out_free_3: |
6aa8b732 | 2823 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2824 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2825 | out_free_2: |
d2308784 | 2826 | out_free_1: |
e9b11c17 | 2827 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2828 | out_free_0a: |
2829 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2830 | out_free_0: |
edba23e5 GN |
2831 | if (fault_page) |
2832 | __free_page(fault_page); | |
bf998156 HY |
2833 | if (hwpoison_page) |
2834 | __free_page(hwpoison_page); | |
d2308784 | 2835 | __free_page(bad_page); |
ca45aaae | 2836 | out: |
f8c16bba | 2837 | kvm_arch_exit(); |
d2308784 | 2838 | out_fail: |
6aa8b732 AK |
2839 | return r; |
2840 | } | |
cb498ea2 | 2841 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2842 | |
cb498ea2 | 2843 | void kvm_exit(void) |
6aa8b732 | 2844 | { |
0ea4ed8e | 2845 | kvm_exit_debug(); |
6aa8b732 | 2846 | misc_deregister(&kvm_dev); |
c16f862d | 2847 | kmem_cache_destroy(kvm_vcpu_cache); |
af585b92 | 2848 | kvm_async_pf_deinit(); |
fb3600cc | 2849 | unregister_syscore_ops(&kvm_syscore_ops); |
6aa8b732 | 2850 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2851 | unregister_cpu_notifier(&kvm_cpu_notifier); |
75b7127c | 2852 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
e9b11c17 | 2853 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2854 | kvm_arch_exit(); |
7f59f492 | 2855 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2856 | __free_page(hwpoison_page); |
cea7bb21 | 2857 | __free_page(bad_page); |
6aa8b732 | 2858 | } |
cb498ea2 | 2859 | EXPORT_SYMBOL_GPL(kvm_exit); |