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