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; |
a36a57b1 TY |
628 | return 0; |
629 | } | |
d48ead8b | 630 | #endif /* !CONFIG_S390 */ |
a36a57b1 | 631 | |
6aa8b732 AK |
632 | /* |
633 | * Allocate some memory and give it an address in the guest physical address | |
634 | * space. | |
635 | * | |
636 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 637 | * |
10589a46 | 638 | * Must be called holding mmap_sem for write. |
6aa8b732 | 639 | */ |
f78e0e2e SY |
640 | int __kvm_set_memory_region(struct kvm *kvm, |
641 | struct kvm_userspace_memory_region *mem, | |
642 | int user_alloc) | |
6aa8b732 | 643 | { |
8234b22e | 644 | int r; |
6aa8b732 | 645 | gfn_t base_gfn; |
28bcb112 HC |
646 | unsigned long npages; |
647 | unsigned long i; | |
6aa8b732 AK |
648 | struct kvm_memory_slot *memslot; |
649 | struct kvm_memory_slot old, new; | |
bc6678a3 | 650 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
651 | |
652 | r = -EINVAL; | |
653 | /* General sanity checks */ | |
654 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
655 | goto out; | |
656 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
657 | goto out; | |
fa3d315a TY |
658 | /* We can read the guest memory with __xxx_user() later on. */ |
659 | if (user_alloc && | |
660 | ((mem->userspace_addr & (PAGE_SIZE - 1)) || | |
9e3bb6b6 HC |
661 | !access_ok(VERIFY_WRITE, |
662 | (void __user *)(unsigned long)mem->userspace_addr, | |
663 | mem->memory_size))) | |
78749809 | 664 | goto out; |
e0d62c7f | 665 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
666 | goto out; |
667 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
668 | goto out; | |
669 | ||
46a26bf5 | 670 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
671 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
672 | npages = mem->memory_size >> PAGE_SHIFT; | |
673 | ||
660c22c4 TY |
674 | r = -EINVAL; |
675 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
676 | goto out; | |
677 | ||
6aa8b732 AK |
678 | if (!npages) |
679 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
680 | ||
6aa8b732 AK |
681 | new = old = *memslot; |
682 | ||
e36d96f7 | 683 | new.id = mem->slot; |
6aa8b732 AK |
684 | new.base_gfn = base_gfn; |
685 | new.npages = npages; | |
686 | new.flags = mem->flags; | |
687 | ||
688 | /* Disallow changing a memory slot's size. */ | |
689 | r = -EINVAL; | |
690 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 691 | goto out_free; |
6aa8b732 AK |
692 | |
693 | /* Check for overlaps */ | |
694 | r = -EEXIST; | |
695 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 696 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 697 | |
4cd481f6 | 698 | if (s == memslot || !s->npages) |
6aa8b732 AK |
699 | continue; |
700 | if (!((base_gfn + npages <= s->base_gfn) || | |
701 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 702 | goto out_free; |
6aa8b732 | 703 | } |
6aa8b732 | 704 | |
6aa8b732 AK |
705 | /* Free page dirty bitmap if unneeded */ |
706 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 707 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
708 | |
709 | r = -ENOMEM; | |
710 | ||
711 | /* Allocate if a slot is being created */ | |
eff0114a | 712 | #ifndef CONFIG_S390 |
8d4e1288 | 713 | if (npages && !new.rmap) { |
26535037 | 714 | new.rmap = vzalloc(npages * sizeof(*new.rmap)); |
290fc38d IE |
715 | |
716 | if (!new.rmap) | |
f78e0e2e | 717 | goto out_free; |
290fc38d | 718 | |
80b14b5b | 719 | new.user_alloc = user_alloc; |
bc6678a3 | 720 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 721 | } |
ec04b260 JR |
722 | if (!npages) |
723 | goto skip_lpage; | |
05da4558 | 724 | |
ec04b260 | 725 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
726 | unsigned long ugfn; |
727 | unsigned long j; | |
728 | int lpages; | |
ec04b260 | 729 | int level = i + 2; |
05da4558 | 730 | |
ec04b260 JR |
731 | /* Avoid unused variable warning if no large pages */ |
732 | (void)level; | |
733 | ||
734 | if (new.lpage_info[i]) | |
735 | continue; | |
736 | ||
82855413 JR |
737 | lpages = 1 + ((base_gfn + npages - 1) |
738 | >> KVM_HPAGE_GFN_SHIFT(level)); | |
739 | lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level); | |
ec04b260 | 740 | |
26535037 | 741 | new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i])); |
ec04b260 JR |
742 | |
743 | if (!new.lpage_info[i]) | |
05da4558 MT |
744 | goto out_free; |
745 | ||
82855413 | 746 | if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 747 | new.lpage_info[i][0].write_count = 1; |
82855413 | 748 | if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 749 | new.lpage_info[i][lpages - 1].write_count = 1; |
ac04527f AK |
750 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
751 | /* | |
752 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
753 | * other, or if explicitly asked to, disable large page |
754 | * support for this slot | |
ac04527f | 755 | */ |
ec04b260 | 756 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 757 | !largepages_enabled) |
ec04b260 JR |
758 | for (j = 0; j < lpages; ++j) |
759 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 760 | } |
6aa8b732 | 761 | |
ec04b260 JR |
762 | skip_lpage: |
763 | ||
6aa8b732 AK |
764 | /* Allocate page dirty bitmap if needed */ |
765 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
a36a57b1 | 766 | if (kvm_create_dirty_bitmap(&new) < 0) |
f78e0e2e | 767 | goto out_free; |
bc6678a3 | 768 | /* destroy any largepage mappings for dirty tracking */ |
6aa8b732 | 769 | } |
3eea8437 CB |
770 | #else /* not defined CONFIG_S390 */ |
771 | new.user_alloc = user_alloc; | |
772 | if (user_alloc) | |
773 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 774 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 775 | |
bc6678a3 MT |
776 | if (!npages) { |
777 | r = -ENOMEM; | |
778 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
779 | if (!slots) | |
780 | goto out_free; | |
781 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
782 | if (mem->slot >= slots->nmemslots) | |
783 | slots->nmemslots = mem->slot + 1; | |
49c7754c | 784 | slots->generation++; |
bc6678a3 MT |
785 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; |
786 | ||
787 | old_memslots = kvm->memslots; | |
788 | rcu_assign_pointer(kvm->memslots, slots); | |
789 | synchronize_srcu_expedited(&kvm->srcu); | |
790 | /* From this point no new shadow pages pointing to a deleted | |
791 | * memslot will be created. | |
792 | * | |
793 | * validation of sp->gfn happens in: | |
794 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
795 | * - kvm_is_visible_gfn (mmu_check_roots) | |
796 | */ | |
34d4cb8f | 797 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
798 | kfree(old_memslots); |
799 | } | |
34d4cb8f | 800 | |
f7784b8e MT |
801 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
802 | if (r) | |
803 | goto out_free; | |
804 | ||
bc6678a3 MT |
805 | /* map the pages in iommu page table */ |
806 | if (npages) { | |
807 | r = kvm_iommu_map_pages(kvm, &new); | |
808 | if (r) | |
809 | goto out_free; | |
810 | } | |
604b38ac | 811 | |
bc6678a3 MT |
812 | r = -ENOMEM; |
813 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
814 | if (!slots) | |
815 | goto out_free; | |
816 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
817 | if (mem->slot >= slots->nmemslots) | |
818 | slots->nmemslots = mem->slot + 1; | |
49c7754c | 819 | slots->generation++; |
bc6678a3 MT |
820 | |
821 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
822 | if (!npages) { | |
823 | new.rmap = NULL; | |
824 | new.dirty_bitmap = NULL; | |
825 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
826 | new.lpage_info[i] = NULL; | |
827 | } | |
828 | ||
829 | slots->memslots[mem->slot] = new; | |
830 | old_memslots = kvm->memslots; | |
831 | rcu_assign_pointer(kvm->memslots, slots); | |
832 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 833 | |
f7784b8e | 834 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 835 | |
ce88decf XG |
836 | /* |
837 | * If the new memory slot is created, we need to clear all | |
838 | * mmio sptes. | |
839 | */ | |
840 | if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT) | |
841 | kvm_arch_flush_shadow(kvm); | |
842 | ||
bc6678a3 MT |
843 | kvm_free_physmem_slot(&old, &new); |
844 | kfree(old_memslots); | |
845 | ||
6aa8b732 AK |
846 | return 0; |
847 | ||
f78e0e2e | 848 | out_free: |
6aa8b732 AK |
849 | kvm_free_physmem_slot(&new, &old); |
850 | out: | |
851 | return r; | |
210c7c4d IE |
852 | |
853 | } | |
f78e0e2e SY |
854 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
855 | ||
856 | int kvm_set_memory_region(struct kvm *kvm, | |
857 | struct kvm_userspace_memory_region *mem, | |
858 | int user_alloc) | |
859 | { | |
860 | int r; | |
861 | ||
79fac95e | 862 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 863 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 864 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
865 | return r; |
866 | } | |
210c7c4d IE |
867 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
868 | ||
1fe779f8 CO |
869 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
870 | struct | |
871 | kvm_userspace_memory_region *mem, | |
872 | int user_alloc) | |
210c7c4d | 873 | { |
e0d62c7f IE |
874 | if (mem->slot >= KVM_MEMORY_SLOTS) |
875 | return -EINVAL; | |
210c7c4d | 876 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
877 | } |
878 | ||
5bb064dc ZX |
879 | int kvm_get_dirty_log(struct kvm *kvm, |
880 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
881 | { |
882 | struct kvm_memory_slot *memslot; | |
883 | int r, i; | |
87bf6e7d | 884 | unsigned long n; |
6aa8b732 AK |
885 | unsigned long any = 0; |
886 | ||
6aa8b732 AK |
887 | r = -EINVAL; |
888 | if (log->slot >= KVM_MEMORY_SLOTS) | |
889 | goto out; | |
890 | ||
46a26bf5 | 891 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
892 | r = -ENOENT; |
893 | if (!memslot->dirty_bitmap) | |
894 | goto out; | |
895 | ||
87bf6e7d | 896 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 897 | |
cd1a4a98 | 898 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
899 | any = memslot->dirty_bitmap[i]; |
900 | ||
901 | r = -EFAULT; | |
902 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
903 | goto out; | |
904 | ||
5bb064dc ZX |
905 | if (any) |
906 | *is_dirty = 1; | |
6aa8b732 AK |
907 | |
908 | r = 0; | |
6aa8b732 | 909 | out: |
6aa8b732 AK |
910 | return r; |
911 | } | |
912 | ||
54dee993 MT |
913 | void kvm_disable_largepages(void) |
914 | { | |
915 | largepages_enabled = false; | |
916 | } | |
917 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
918 | ||
cea7bb21 IE |
919 | int is_error_page(struct page *page) |
920 | { | |
edba23e5 | 921 | return page == bad_page || page == hwpoison_page || page == fault_page; |
cea7bb21 IE |
922 | } |
923 | EXPORT_SYMBOL_GPL(is_error_page); | |
924 | ||
35149e21 AL |
925 | int is_error_pfn(pfn_t pfn) |
926 | { | |
edba23e5 | 927 | return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn; |
35149e21 AL |
928 | } |
929 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
930 | ||
bf998156 HY |
931 | int is_hwpoison_pfn(pfn_t pfn) |
932 | { | |
933 | return pfn == hwpoison_pfn; | |
934 | } | |
935 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
936 | ||
edba23e5 GN |
937 | int is_fault_pfn(pfn_t pfn) |
938 | { | |
939 | return pfn == fault_pfn; | |
940 | } | |
941 | EXPORT_SYMBOL_GPL(is_fault_pfn); | |
942 | ||
fce92dce XG |
943 | int is_noslot_pfn(pfn_t pfn) |
944 | { | |
945 | return pfn == bad_pfn; | |
946 | } | |
947 | EXPORT_SYMBOL_GPL(is_noslot_pfn); | |
948 | ||
949 | int is_invalid_pfn(pfn_t pfn) | |
950 | { | |
951 | return pfn == hwpoison_pfn || pfn == fault_pfn; | |
952 | } | |
953 | EXPORT_SYMBOL_GPL(is_invalid_pfn); | |
954 | ||
f9d46eb0 IE |
955 | static inline unsigned long bad_hva(void) |
956 | { | |
957 | return PAGE_OFFSET; | |
958 | } | |
959 | ||
960 | int kvm_is_error_hva(unsigned long addr) | |
961 | { | |
962 | return addr == bad_hva(); | |
963 | } | |
964 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
965 | ||
49c7754c GN |
966 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots, |
967 | gfn_t gfn) | |
6aa8b732 AK |
968 | { |
969 | int i; | |
970 | ||
46a26bf5 MT |
971 | for (i = 0; i < slots->nmemslots; ++i) { |
972 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
973 | |
974 | if (gfn >= memslot->base_gfn | |
975 | && gfn < memslot->base_gfn + memslot->npages) | |
976 | return memslot; | |
977 | } | |
8b6d44c7 | 978 | return NULL; |
6aa8b732 | 979 | } |
49c7754c GN |
980 | |
981 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
982 | { | |
983 | return __gfn_to_memslot(kvm_memslots(kvm), gfn); | |
984 | } | |
a1f4d395 | 985 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 986 | |
e0d62c7f IE |
987 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
988 | { | |
989 | int i; | |
90d83dc3 | 990 | struct kvm_memslots *slots = kvm_memslots(kvm); |
e0d62c7f | 991 | |
e0d62c7f | 992 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 993 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 994 | |
bc6678a3 MT |
995 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
996 | continue; | |
997 | ||
e0d62c7f IE |
998 | if (gfn >= memslot->base_gfn |
999 | && gfn < memslot->base_gfn + memslot->npages) | |
1000 | return 1; | |
1001 | } | |
1002 | return 0; | |
1003 | } | |
1004 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
1005 | ||
8f0b1ab6 JR |
1006 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
1007 | { | |
1008 | struct vm_area_struct *vma; | |
1009 | unsigned long addr, size; | |
1010 | ||
1011 | size = PAGE_SIZE; | |
1012 | ||
1013 | addr = gfn_to_hva(kvm, gfn); | |
1014 | if (kvm_is_error_hva(addr)) | |
1015 | return PAGE_SIZE; | |
1016 | ||
1017 | down_read(¤t->mm->mmap_sem); | |
1018 | vma = find_vma(current->mm, addr); | |
1019 | if (!vma) | |
1020 | goto out; | |
1021 | ||
1022 | size = vma_kernel_pagesize(vma); | |
1023 | ||
1024 | out: | |
1025 | up_read(¤t->mm->mmap_sem); | |
1026 | ||
1027 | return size; | |
1028 | } | |
1029 | ||
49c7754c | 1030 | static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn, |
48987781 | 1031 | gfn_t *nr_pages) |
539cb660 | 1032 | { |
bc6678a3 | 1033 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 1034 | return bad_hva(); |
48987781 XG |
1035 | |
1036 | if (nr_pages) | |
1037 | *nr_pages = slot->npages - (gfn - slot->base_gfn); | |
1038 | ||
f5c98031 | 1039 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 1040 | } |
48987781 XG |
1041 | |
1042 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) | |
1043 | { | |
49c7754c | 1044 | return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL); |
48987781 | 1045 | } |
0d150298 | 1046 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1047 | |
8030089f GN |
1048 | static pfn_t get_fault_pfn(void) |
1049 | { | |
1050 | get_page(fault_page); | |
1051 | return fault_pfn; | |
1052 | } | |
1053 | ||
0857b9e9 GN |
1054 | int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm, |
1055 | unsigned long start, int write, struct page **page) | |
1056 | { | |
1057 | int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET; | |
1058 | ||
1059 | if (write) | |
1060 | flags |= FOLL_WRITE; | |
1061 | ||
1062 | return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL); | |
1063 | } | |
1064 | ||
fafc3dba HY |
1065 | static inline int check_user_page_hwpoison(unsigned long addr) |
1066 | { | |
1067 | int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE; | |
1068 | ||
1069 | rc = __get_user_pages(current, current->mm, addr, 1, | |
1070 | flags, NULL, NULL, NULL); | |
1071 | return rc == -EHWPOISON; | |
1072 | } | |
1073 | ||
af585b92 | 1074 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic, |
612819c3 | 1075 | bool *async, bool write_fault, bool *writable) |
954bbbc2 | 1076 | { |
8d4e1288 | 1077 | struct page *page[1]; |
af585b92 | 1078 | int npages = 0; |
2e2e3738 | 1079 | pfn_t pfn; |
954bbbc2 | 1080 | |
af585b92 GN |
1081 | /* we can do it either atomically or asynchronously, not both */ |
1082 | BUG_ON(atomic && async); | |
1083 | ||
612819c3 MT |
1084 | BUG_ON(!write_fault && !writable); |
1085 | ||
1086 | if (writable) | |
1087 | *writable = true; | |
1088 | ||
af585b92 | 1089 | if (atomic || async) |
887c08ac | 1090 | npages = __get_user_pages_fast(addr, 1, 1, page); |
af585b92 GN |
1091 | |
1092 | if (unlikely(npages != 1) && !atomic) { | |
887c08ac | 1093 | might_sleep(); |
612819c3 MT |
1094 | |
1095 | if (writable) | |
1096 | *writable = write_fault; | |
1097 | ||
0857b9e9 GN |
1098 | if (async) { |
1099 | down_read(¤t->mm->mmap_sem); | |
1100 | npages = get_user_page_nowait(current, current->mm, | |
1101 | addr, write_fault, page); | |
1102 | up_read(¤t->mm->mmap_sem); | |
1103 | } else | |
1104 | npages = get_user_pages_fast(addr, 1, write_fault, | |
1105 | page); | |
612819c3 MT |
1106 | |
1107 | /* map read fault as writable if possible */ | |
1108 | if (unlikely(!write_fault) && npages == 1) { | |
1109 | struct page *wpage[1]; | |
1110 | ||
1111 | npages = __get_user_pages_fast(addr, 1, 1, wpage); | |
1112 | if (npages == 1) { | |
1113 | *writable = true; | |
1114 | put_page(page[0]); | |
1115 | page[0] = wpage[0]; | |
1116 | } | |
1117 | npages = 1; | |
1118 | } | |
887c08ac | 1119 | } |
539cb660 | 1120 | |
2e2e3738 AL |
1121 | if (unlikely(npages != 1)) { |
1122 | struct vm_area_struct *vma; | |
1123 | ||
887c08ac | 1124 | if (atomic) |
8030089f | 1125 | return get_fault_pfn(); |
887c08ac | 1126 | |
bbeb3406 | 1127 | down_read(¤t->mm->mmap_sem); |
0857b9e9 GN |
1128 | if (npages == -EHWPOISON || |
1129 | (!async && check_user_page_hwpoison(addr))) { | |
bbeb3406 | 1130 | up_read(¤t->mm->mmap_sem); |
bf998156 HY |
1131 | get_page(hwpoison_page); |
1132 | return page_to_pfn(hwpoison_page); | |
1133 | } | |
1134 | ||
8030089f | 1135 | vma = find_vma_intersection(current->mm, addr, addr+1); |
4c2155ce | 1136 | |
8030089f GN |
1137 | if (vma == NULL) |
1138 | pfn = get_fault_pfn(); | |
1139 | else if ((vma->vm_flags & VM_PFNMAP)) { | |
1140 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + | |
1141 | vma->vm_pgoff; | |
1142 | BUG_ON(!kvm_is_mmio_pfn(pfn)); | |
1143 | } else { | |
1144 | if (async && (vma->vm_flags & VM_WRITE)) | |
af585b92 | 1145 | *async = true; |
8030089f | 1146 | pfn = get_fault_pfn(); |
2e2e3738 | 1147 | } |
4c2155ce | 1148 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1149 | } else |
1150 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1151 | |
2e2e3738 | 1152 | return pfn; |
35149e21 AL |
1153 | } |
1154 | ||
887c08ac XG |
1155 | pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr) |
1156 | { | |
612819c3 | 1157 | return hva_to_pfn(kvm, addr, true, NULL, true, NULL); |
887c08ac XG |
1158 | } |
1159 | EXPORT_SYMBOL_GPL(hva_to_pfn_atomic); | |
1160 | ||
612819c3 MT |
1161 | static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async, |
1162 | bool write_fault, bool *writable) | |
506f0d6f MT |
1163 | { |
1164 | unsigned long addr; | |
1165 | ||
af585b92 GN |
1166 | if (async) |
1167 | *async = false; | |
1168 | ||
506f0d6f MT |
1169 | addr = gfn_to_hva(kvm, gfn); |
1170 | if (kvm_is_error_hva(addr)) { | |
1171 | get_page(bad_page); | |
1172 | return page_to_pfn(bad_page); | |
1173 | } | |
1174 | ||
612819c3 | 1175 | return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable); |
365fb3fd XG |
1176 | } |
1177 | ||
1178 | pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn) | |
1179 | { | |
612819c3 | 1180 | return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL); |
365fb3fd XG |
1181 | } |
1182 | EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic); | |
1183 | ||
612819c3 MT |
1184 | pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async, |
1185 | bool write_fault, bool *writable) | |
af585b92 | 1186 | { |
612819c3 | 1187 | return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable); |
af585b92 GN |
1188 | } |
1189 | EXPORT_SYMBOL_GPL(gfn_to_pfn_async); | |
1190 | ||
365fb3fd XG |
1191 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
1192 | { | |
612819c3 | 1193 | return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL); |
506f0d6f | 1194 | } |
35149e21 AL |
1195 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
1196 | ||
612819c3 MT |
1197 | pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, |
1198 | bool *writable) | |
1199 | { | |
1200 | return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable); | |
1201 | } | |
1202 | EXPORT_SYMBOL_GPL(gfn_to_pfn_prot); | |
1203 | ||
506f0d6f MT |
1204 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
1205 | struct kvm_memory_slot *slot, gfn_t gfn) | |
1206 | { | |
1207 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
612819c3 | 1208 | return hva_to_pfn(kvm, addr, false, NULL, true, NULL); |
506f0d6f MT |
1209 | } |
1210 | ||
48987781 XG |
1211 | int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages, |
1212 | int nr_pages) | |
1213 | { | |
1214 | unsigned long addr; | |
1215 | gfn_t entry; | |
1216 | ||
49c7754c | 1217 | addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry); |
48987781 XG |
1218 | if (kvm_is_error_hva(addr)) |
1219 | return -1; | |
1220 | ||
1221 | if (entry < nr_pages) | |
1222 | return 0; | |
1223 | ||
1224 | return __get_user_pages_fast(addr, nr_pages, 1, pages); | |
1225 | } | |
1226 | EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic); | |
1227 | ||
35149e21 AL |
1228 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1229 | { | |
2e2e3738 AL |
1230 | pfn_t pfn; |
1231 | ||
1232 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1233 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1234 | return pfn_to_page(pfn); |
1235 | ||
c77fb9dc | 1236 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1237 | |
1238 | get_page(bad_page); | |
1239 | return bad_page; | |
954bbbc2 | 1240 | } |
aab61cc0 | 1241 | |
954bbbc2 AK |
1242 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1243 | ||
b4231d61 IE |
1244 | void kvm_release_page_clean(struct page *page) |
1245 | { | |
35149e21 | 1246 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1247 | } |
1248 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1249 | ||
35149e21 AL |
1250 | void kvm_release_pfn_clean(pfn_t pfn) |
1251 | { | |
c77fb9dc | 1252 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1253 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1254 | } |
1255 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1256 | ||
b4231d61 | 1257 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1258 | { |
35149e21 AL |
1259 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1260 | } | |
1261 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1262 | ||
1263 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1264 | { | |
1265 | kvm_set_pfn_dirty(pfn); | |
1266 | kvm_release_pfn_clean(pfn); | |
1267 | } | |
1268 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1269 | ||
1270 | void kvm_set_page_dirty(struct page *page) | |
1271 | { | |
1272 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1273 | } | |
1274 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1275 | ||
1276 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1277 | { | |
c77fb9dc | 1278 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1279 | struct page *page = pfn_to_page(pfn); |
1280 | if (!PageReserved(page)) | |
1281 | SetPageDirty(page); | |
1282 | } | |
8a7ae055 | 1283 | } |
35149e21 AL |
1284 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1285 | ||
1286 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1287 | { | |
c77fb9dc | 1288 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1289 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1290 | } |
1291 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1292 | ||
1293 | void kvm_get_pfn(pfn_t pfn) | |
1294 | { | |
c77fb9dc | 1295 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1296 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1297 | } |
1298 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1299 | |
195aefde IE |
1300 | static int next_segment(unsigned long len, int offset) |
1301 | { | |
1302 | if (len > PAGE_SIZE - offset) | |
1303 | return PAGE_SIZE - offset; | |
1304 | else | |
1305 | return len; | |
1306 | } | |
1307 | ||
1308 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1309 | int len) | |
1310 | { | |
e0506bcb IE |
1311 | int r; |
1312 | unsigned long addr; | |
195aefde | 1313 | |
e0506bcb IE |
1314 | addr = gfn_to_hva(kvm, gfn); |
1315 | if (kvm_is_error_hva(addr)) | |
1316 | return -EFAULT; | |
fa3d315a | 1317 | r = __copy_from_user(data, (void __user *)addr + offset, len); |
e0506bcb | 1318 | if (r) |
195aefde | 1319 | return -EFAULT; |
195aefde IE |
1320 | return 0; |
1321 | } | |
1322 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1323 | ||
1324 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1325 | { | |
1326 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1327 | int seg; | |
1328 | int offset = offset_in_page(gpa); | |
1329 | int ret; | |
1330 | ||
1331 | while ((seg = next_segment(len, offset)) != 0) { | |
1332 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1333 | if (ret < 0) | |
1334 | return ret; | |
1335 | offset = 0; | |
1336 | len -= seg; | |
1337 | data += seg; | |
1338 | ++gfn; | |
1339 | } | |
1340 | return 0; | |
1341 | } | |
1342 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1343 | ||
7ec54588 MT |
1344 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1345 | unsigned long len) | |
1346 | { | |
1347 | int r; | |
1348 | unsigned long addr; | |
1349 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1350 | int offset = offset_in_page(gpa); | |
1351 | ||
1352 | addr = gfn_to_hva(kvm, gfn); | |
1353 | if (kvm_is_error_hva(addr)) | |
1354 | return -EFAULT; | |
0aac03f0 | 1355 | pagefault_disable(); |
7ec54588 | 1356 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1357 | pagefault_enable(); |
7ec54588 MT |
1358 | if (r) |
1359 | return -EFAULT; | |
1360 | return 0; | |
1361 | } | |
1362 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1363 | ||
195aefde IE |
1364 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1365 | int offset, int len) | |
1366 | { | |
e0506bcb IE |
1367 | int r; |
1368 | unsigned long addr; | |
195aefde | 1369 | |
e0506bcb IE |
1370 | addr = gfn_to_hva(kvm, gfn); |
1371 | if (kvm_is_error_hva(addr)) | |
1372 | return -EFAULT; | |
8b0cedff | 1373 | r = __copy_to_user((void __user *)addr + offset, data, len); |
e0506bcb | 1374 | if (r) |
195aefde | 1375 | return -EFAULT; |
195aefde IE |
1376 | mark_page_dirty(kvm, gfn); |
1377 | return 0; | |
1378 | } | |
1379 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1380 | ||
1381 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1382 | unsigned long len) | |
1383 | { | |
1384 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1385 | int seg; | |
1386 | int offset = offset_in_page(gpa); | |
1387 | int ret; | |
1388 | ||
1389 | while ((seg = next_segment(len, offset)) != 0) { | |
1390 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1391 | if (ret < 0) | |
1392 | return ret; | |
1393 | offset = 0; | |
1394 | len -= seg; | |
1395 | data += seg; | |
1396 | ++gfn; | |
1397 | } | |
1398 | return 0; | |
1399 | } | |
1400 | ||
49c7754c GN |
1401 | int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1402 | gpa_t gpa) | |
1403 | { | |
1404 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1405 | int offset = offset_in_page(gpa); | |
1406 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1407 | ||
1408 | ghc->gpa = gpa; | |
1409 | ghc->generation = slots->generation; | |
1410 | ghc->memslot = __gfn_to_memslot(slots, gfn); | |
1411 | ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL); | |
1412 | if (!kvm_is_error_hva(ghc->hva)) | |
1413 | ghc->hva += offset; | |
1414 | else | |
1415 | return -EFAULT; | |
1416 | ||
1417 | return 0; | |
1418 | } | |
1419 | EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init); | |
1420 | ||
1421 | int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, | |
1422 | void *data, unsigned long len) | |
1423 | { | |
1424 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1425 | int r; | |
1426 | ||
1427 | if (slots->generation != ghc->generation) | |
1428 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1429 | ||
1430 | if (kvm_is_error_hva(ghc->hva)) | |
1431 | return -EFAULT; | |
1432 | ||
8b0cedff | 1433 | r = __copy_to_user((void __user *)ghc->hva, data, len); |
49c7754c GN |
1434 | if (r) |
1435 | return -EFAULT; | |
1436 | mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT); | |
1437 | ||
1438 | return 0; | |
1439 | } | |
1440 | EXPORT_SYMBOL_GPL(kvm_write_guest_cached); | |
1441 | ||
e03b644f GN |
1442 | int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1443 | void *data, unsigned long len) | |
1444 | { | |
1445 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1446 | int r; | |
1447 | ||
1448 | if (slots->generation != ghc->generation) | |
1449 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1450 | ||
1451 | if (kvm_is_error_hva(ghc->hva)) | |
1452 | return -EFAULT; | |
1453 | ||
1454 | r = __copy_from_user(data, (void __user *)ghc->hva, len); | |
1455 | if (r) | |
1456 | return -EFAULT; | |
1457 | ||
1458 | return 0; | |
1459 | } | |
1460 | EXPORT_SYMBOL_GPL(kvm_read_guest_cached); | |
1461 | ||
195aefde IE |
1462 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) |
1463 | { | |
3bcc8a8c HC |
1464 | return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page, |
1465 | offset, len); | |
195aefde IE |
1466 | } |
1467 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1468 | ||
1469 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1470 | { | |
1471 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1472 | int seg; | |
1473 | int offset = offset_in_page(gpa); | |
1474 | int ret; | |
1475 | ||
1476 | while ((seg = next_segment(len, offset)) != 0) { | |
1477 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1478 | if (ret < 0) | |
1479 | return ret; | |
1480 | offset = 0; | |
1481 | len -= seg; | |
1482 | ++gfn; | |
1483 | } | |
1484 | return 0; | |
1485 | } | |
1486 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1487 | ||
49c7754c GN |
1488 | void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, |
1489 | gfn_t gfn) | |
6aa8b732 | 1490 | { |
7e9d619d RR |
1491 | if (memslot && memslot->dirty_bitmap) { |
1492 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1493 | |
cd7e48c5 | 1494 | __set_bit_le(rel_gfn, memslot->dirty_bitmap); |
6aa8b732 AK |
1495 | } |
1496 | } | |
1497 | ||
49c7754c GN |
1498 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1499 | { | |
1500 | struct kvm_memory_slot *memslot; | |
1501 | ||
1502 | memslot = gfn_to_memslot(kvm, gfn); | |
1503 | mark_page_dirty_in_slot(kvm, memslot, gfn); | |
1504 | } | |
1505 | ||
b6958ce4 ED |
1506 | /* |
1507 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1508 | */ | |
8776e519 | 1509 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1510 | { |
e5c239cf MT |
1511 | DEFINE_WAIT(wait); |
1512 | ||
1513 | for (;;) { | |
1514 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1515 | ||
a1b37100 | 1516 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1517 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1518 | break; |
d7690175 | 1519 | } |
09cec754 GN |
1520 | if (kvm_cpu_has_pending_timer(vcpu)) |
1521 | break; | |
e5c239cf MT |
1522 | if (signal_pending(current)) |
1523 | break; | |
1524 | ||
b6958ce4 | 1525 | schedule(); |
b6958ce4 | 1526 | } |
d3bef15f | 1527 | |
e5c239cf | 1528 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1529 | } |
1530 | ||
6aa8b732 AK |
1531 | void kvm_resched(struct kvm_vcpu *vcpu) |
1532 | { | |
3fca0365 YD |
1533 | if (!need_resched()) |
1534 | return; | |
6aa8b732 | 1535 | cond_resched(); |
6aa8b732 AK |
1536 | } |
1537 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1538 | ||
217ece61 | 1539 | void kvm_vcpu_on_spin(struct kvm_vcpu *me) |
d255f4f2 | 1540 | { |
217ece61 RR |
1541 | struct kvm *kvm = me->kvm; |
1542 | struct kvm_vcpu *vcpu; | |
1543 | int last_boosted_vcpu = me->kvm->last_boosted_vcpu; | |
1544 | int yielded = 0; | |
1545 | int pass; | |
1546 | int i; | |
d255f4f2 | 1547 | |
217ece61 RR |
1548 | /* |
1549 | * We boost the priority of a VCPU that is runnable but not | |
1550 | * currently running, because it got preempted by something | |
1551 | * else and called schedule in __vcpu_run. Hopefully that | |
1552 | * VCPU is holding the lock that we need and will release it. | |
1553 | * We approximate round-robin by starting at the last boosted VCPU. | |
1554 | */ | |
1555 | for (pass = 0; pass < 2 && !yielded; pass++) { | |
1556 | kvm_for_each_vcpu(i, vcpu, kvm) { | |
1557 | struct task_struct *task = NULL; | |
1558 | struct pid *pid; | |
1559 | if (!pass && i < last_boosted_vcpu) { | |
1560 | i = last_boosted_vcpu; | |
1561 | continue; | |
1562 | } else if (pass && i > last_boosted_vcpu) | |
1563 | break; | |
1564 | if (vcpu == me) | |
1565 | continue; | |
1566 | if (waitqueue_active(&vcpu->wq)) | |
1567 | continue; | |
1568 | rcu_read_lock(); | |
1569 | pid = rcu_dereference(vcpu->pid); | |
1570 | if (pid) | |
1571 | task = get_pid_task(vcpu->pid, PIDTYPE_PID); | |
1572 | rcu_read_unlock(); | |
1573 | if (!task) | |
1574 | continue; | |
1575 | if (task->flags & PF_VCPU) { | |
1576 | put_task_struct(task); | |
1577 | continue; | |
1578 | } | |
1579 | if (yield_to(task, 1)) { | |
1580 | put_task_struct(task); | |
1581 | kvm->last_boosted_vcpu = i; | |
1582 | yielded = 1; | |
1583 | break; | |
1584 | } | |
1585 | put_task_struct(task); | |
1586 | } | |
1587 | } | |
d255f4f2 ZE |
1588 | } |
1589 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1590 | ||
e4a533a4 | 1591 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1592 | { |
1593 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1594 | struct page *page; |
1595 | ||
e4a533a4 | 1596 | if (vmf->pgoff == 0) |
039576c0 | 1597 | page = virt_to_page(vcpu->run); |
09566765 | 1598 | #ifdef CONFIG_X86 |
e4a533a4 | 1599 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1600 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1601 | #endif |
1602 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1603 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1604 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1605 | #endif |
039576c0 | 1606 | else |
e4a533a4 | 1607 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1608 | get_page(page); |
e4a533a4 | 1609 | vmf->page = page; |
1610 | return 0; | |
9a2bb7f4 AK |
1611 | } |
1612 | ||
f0f37e2f | 1613 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1614 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1615 | }; |
1616 | ||
1617 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1618 | { | |
1619 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1620 | return 0; | |
1621 | } | |
1622 | ||
bccf2150 AK |
1623 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1624 | { | |
1625 | struct kvm_vcpu *vcpu = filp->private_data; | |
1626 | ||
66c0b394 | 1627 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1628 | return 0; |
1629 | } | |
1630 | ||
3d3aab1b | 1631 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1632 | .release = kvm_vcpu_release, |
1633 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1dda606c AG |
1634 | #ifdef CONFIG_COMPAT |
1635 | .compat_ioctl = kvm_vcpu_compat_ioctl, | |
1636 | #endif | |
9a2bb7f4 | 1637 | .mmap = kvm_vcpu_mmap, |
6038f373 | 1638 | .llseek = noop_llseek, |
bccf2150 AK |
1639 | }; |
1640 | ||
1641 | /* | |
1642 | * Allocates an inode for the vcpu. | |
1643 | */ | |
1644 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1645 | { | |
628ff7c1 | 1646 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1647 | } |
1648 | ||
c5ea7660 AK |
1649 | /* |
1650 | * Creates some virtual cpus. Good luck creating more than one. | |
1651 | */ | |
73880c80 | 1652 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1653 | { |
1654 | int r; | |
988a2cae | 1655 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1656 | |
73880c80 | 1657 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1658 | if (IS_ERR(vcpu)) |
1659 | return PTR_ERR(vcpu); | |
c5ea7660 | 1660 | |
15ad7146 AK |
1661 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1662 | ||
26e5215f AK |
1663 | r = kvm_arch_vcpu_setup(vcpu); |
1664 | if (r) | |
d780592b | 1665 | goto vcpu_destroy; |
26e5215f | 1666 | |
11ec2804 | 1667 | mutex_lock(&kvm->lock); |
73880c80 GN |
1668 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1669 | r = -EINVAL; | |
d780592b | 1670 | goto unlock_vcpu_destroy; |
fb3f0f51 | 1671 | } |
73880c80 | 1672 | |
988a2cae GN |
1673 | kvm_for_each_vcpu(r, v, kvm) |
1674 | if (v->vcpu_id == id) { | |
73880c80 | 1675 | r = -EEXIST; |
d780592b | 1676 | goto unlock_vcpu_destroy; |
73880c80 GN |
1677 | } |
1678 | ||
1679 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1680 | |
fb3f0f51 | 1681 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1682 | kvm_get_kvm(kvm); |
bccf2150 | 1683 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1684 | if (r < 0) { |
1685 | kvm_put_kvm(kvm); | |
d780592b | 1686 | goto unlock_vcpu_destroy; |
73880c80 GN |
1687 | } |
1688 | ||
1689 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1690 | smp_wmb(); | |
1691 | atomic_inc(&kvm->online_vcpus); | |
1692 | ||
1693 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1694 | if (kvm->bsp_vcpu_id == id) | |
1695 | kvm->bsp_vcpu = vcpu; | |
1696 | #endif | |
1697 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1698 | return r; |
39c3b86e | 1699 | |
d780592b | 1700 | unlock_vcpu_destroy: |
7d8fece6 | 1701 | mutex_unlock(&kvm->lock); |
d780592b | 1702 | vcpu_destroy: |
d40ccc62 | 1703 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1704 | return r; |
1705 | } | |
1706 | ||
1961d276 AK |
1707 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1708 | { | |
1709 | if (sigset) { | |
1710 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1711 | vcpu->sigset_active = 1; | |
1712 | vcpu->sigset = *sigset; | |
1713 | } else | |
1714 | vcpu->sigset_active = 0; | |
1715 | return 0; | |
1716 | } | |
1717 | ||
bccf2150 AK |
1718 | static long kvm_vcpu_ioctl(struct file *filp, |
1719 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1720 | { |
bccf2150 | 1721 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1722 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1723 | int r; |
fa3795a7 DH |
1724 | struct kvm_fpu *fpu = NULL; |
1725 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1726 | |
6d4e4c4f AK |
1727 | if (vcpu->kvm->mm != current->mm) |
1728 | return -EIO; | |
2122ff5e AK |
1729 | |
1730 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1731 | /* | |
1732 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1733 | * so vcpu_load() would break it. | |
1734 | */ | |
1735 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1736 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1737 | #endif | |
1738 | ||
1739 | ||
1740 | vcpu_load(vcpu); | |
6aa8b732 | 1741 | switch (ioctl) { |
9a2bb7f4 | 1742 | case KVM_RUN: |
f0fe5108 AK |
1743 | r = -EINVAL; |
1744 | if (arg) | |
1745 | goto out; | |
b6c7a5dc | 1746 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
64be5007 | 1747 | trace_kvm_userspace_exit(vcpu->run->exit_reason, r); |
6aa8b732 | 1748 | break; |
6aa8b732 | 1749 | case KVM_GET_REGS: { |
3e4bb3ac | 1750 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1751 | |
3e4bb3ac XZ |
1752 | r = -ENOMEM; |
1753 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1754 | if (!kvm_regs) | |
6aa8b732 | 1755 | goto out; |
3e4bb3ac XZ |
1756 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1757 | if (r) | |
1758 | goto out_free1; | |
6aa8b732 | 1759 | r = -EFAULT; |
3e4bb3ac XZ |
1760 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1761 | goto out_free1; | |
6aa8b732 | 1762 | r = 0; |
3e4bb3ac XZ |
1763 | out_free1: |
1764 | kfree(kvm_regs); | |
6aa8b732 AK |
1765 | break; |
1766 | } | |
1767 | case KVM_SET_REGS: { | |
3e4bb3ac | 1768 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1769 | |
3e4bb3ac XZ |
1770 | r = -ENOMEM; |
1771 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1772 | if (!kvm_regs) | |
6aa8b732 | 1773 | goto out; |
3e4bb3ac XZ |
1774 | r = -EFAULT; |
1775 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1776 | goto out_free2; | |
1777 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1778 | if (r) |
3e4bb3ac | 1779 | goto out_free2; |
6aa8b732 | 1780 | r = 0; |
3e4bb3ac XZ |
1781 | out_free2: |
1782 | kfree(kvm_regs); | |
6aa8b732 AK |
1783 | break; |
1784 | } | |
1785 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1786 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1787 | r = -ENOMEM; | |
1788 | if (!kvm_sregs) | |
1789 | goto out; | |
1790 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1791 | if (r) |
1792 | goto out; | |
1793 | r = -EFAULT; | |
fa3795a7 | 1794 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1795 | goto out; |
1796 | r = 0; | |
1797 | break; | |
1798 | } | |
1799 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1800 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1801 | r = -ENOMEM; | |
1802 | if (!kvm_sregs) | |
1803 | goto out; | |
6aa8b732 | 1804 | r = -EFAULT; |
fa3795a7 | 1805 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1806 | goto out; |
fa3795a7 | 1807 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1808 | if (r) |
1809 | goto out; | |
1810 | r = 0; | |
1811 | break; | |
1812 | } | |
62d9f0db MT |
1813 | case KVM_GET_MP_STATE: { |
1814 | struct kvm_mp_state mp_state; | |
1815 | ||
1816 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1817 | if (r) | |
1818 | goto out; | |
1819 | r = -EFAULT; | |
1820 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1821 | goto out; | |
1822 | r = 0; | |
1823 | break; | |
1824 | } | |
1825 | case KVM_SET_MP_STATE: { | |
1826 | struct kvm_mp_state mp_state; | |
1827 | ||
1828 | r = -EFAULT; | |
1829 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1830 | goto out; | |
1831 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1832 | if (r) | |
1833 | goto out; | |
1834 | r = 0; | |
1835 | break; | |
1836 | } | |
6aa8b732 AK |
1837 | case KVM_TRANSLATE: { |
1838 | struct kvm_translation tr; | |
1839 | ||
1840 | r = -EFAULT; | |
2f366987 | 1841 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1842 | goto out; |
8b006791 | 1843 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1844 | if (r) |
1845 | goto out; | |
1846 | r = -EFAULT; | |
2f366987 | 1847 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1848 | goto out; |
1849 | r = 0; | |
1850 | break; | |
1851 | } | |
d0bfb940 JK |
1852 | case KVM_SET_GUEST_DEBUG: { |
1853 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1854 | |
1855 | r = -EFAULT; | |
2f366987 | 1856 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1857 | goto out; |
d0bfb940 | 1858 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1859 | if (r) |
1860 | goto out; | |
1861 | r = 0; | |
1862 | break; | |
1863 | } | |
1961d276 AK |
1864 | case KVM_SET_SIGNAL_MASK: { |
1865 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1866 | struct kvm_signal_mask kvm_sigmask; | |
1867 | sigset_t sigset, *p; | |
1868 | ||
1869 | p = NULL; | |
1870 | if (argp) { | |
1871 | r = -EFAULT; | |
1872 | if (copy_from_user(&kvm_sigmask, argp, | |
1873 | sizeof kvm_sigmask)) | |
1874 | goto out; | |
1875 | r = -EINVAL; | |
1876 | if (kvm_sigmask.len != sizeof sigset) | |
1877 | goto out; | |
1878 | r = -EFAULT; | |
1879 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1880 | sizeof sigset)) | |
1881 | goto out; | |
1882 | p = &sigset; | |
1883 | } | |
376d41ff | 1884 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1885 | break; |
1886 | } | |
b8836737 | 1887 | case KVM_GET_FPU: { |
fa3795a7 DH |
1888 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1889 | r = -ENOMEM; | |
1890 | if (!fpu) | |
1891 | goto out; | |
1892 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1893 | if (r) |
1894 | goto out; | |
1895 | r = -EFAULT; | |
fa3795a7 | 1896 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1897 | goto out; |
1898 | r = 0; | |
1899 | break; | |
1900 | } | |
1901 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1902 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1903 | r = -ENOMEM; | |
1904 | if (!fpu) | |
1905 | goto out; | |
b8836737 | 1906 | r = -EFAULT; |
fa3795a7 | 1907 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1908 | goto out; |
fa3795a7 | 1909 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1910 | if (r) |
1911 | goto out; | |
1912 | r = 0; | |
1913 | break; | |
1914 | } | |
bccf2150 | 1915 | default: |
313a3dc7 | 1916 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1917 | } |
1918 | out: | |
2122ff5e | 1919 | vcpu_put(vcpu); |
fa3795a7 DH |
1920 | kfree(fpu); |
1921 | kfree(kvm_sregs); | |
bccf2150 AK |
1922 | return r; |
1923 | } | |
1924 | ||
1dda606c AG |
1925 | #ifdef CONFIG_COMPAT |
1926 | static long kvm_vcpu_compat_ioctl(struct file *filp, | |
1927 | unsigned int ioctl, unsigned long arg) | |
1928 | { | |
1929 | struct kvm_vcpu *vcpu = filp->private_data; | |
1930 | void __user *argp = compat_ptr(arg); | |
1931 | int r; | |
1932 | ||
1933 | if (vcpu->kvm->mm != current->mm) | |
1934 | return -EIO; | |
1935 | ||
1936 | switch (ioctl) { | |
1937 | case KVM_SET_SIGNAL_MASK: { | |
1938 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1939 | struct kvm_signal_mask kvm_sigmask; | |
1940 | compat_sigset_t csigset; | |
1941 | sigset_t sigset; | |
1942 | ||
1943 | if (argp) { | |
1944 | r = -EFAULT; | |
1945 | if (copy_from_user(&kvm_sigmask, argp, | |
1946 | sizeof kvm_sigmask)) | |
1947 | goto out; | |
1948 | r = -EINVAL; | |
1949 | if (kvm_sigmask.len != sizeof csigset) | |
1950 | goto out; | |
1951 | r = -EFAULT; | |
1952 | if (copy_from_user(&csigset, sigmask_arg->sigset, | |
1953 | sizeof csigset)) | |
1954 | goto out; | |
1955 | } | |
1956 | sigset_from_compat(&sigset, &csigset); | |
1957 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
1958 | break; | |
1959 | } | |
1960 | default: | |
1961 | r = kvm_vcpu_ioctl(filp, ioctl, arg); | |
1962 | } | |
1963 | ||
1964 | out: | |
1965 | return r; | |
1966 | } | |
1967 | #endif | |
1968 | ||
bccf2150 AK |
1969 | static long kvm_vm_ioctl(struct file *filp, |
1970 | unsigned int ioctl, unsigned long arg) | |
1971 | { | |
1972 | struct kvm *kvm = filp->private_data; | |
1973 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1974 | int r; |
bccf2150 | 1975 | |
6d4e4c4f AK |
1976 | if (kvm->mm != current->mm) |
1977 | return -EIO; | |
bccf2150 AK |
1978 | switch (ioctl) { |
1979 | case KVM_CREATE_VCPU: | |
1980 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1981 | if (r < 0) | |
1982 | goto out; | |
1983 | break; | |
6fc138d2 IE |
1984 | case KVM_SET_USER_MEMORY_REGION: { |
1985 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1986 | ||
1987 | r = -EFAULT; | |
1988 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1989 | sizeof kvm_userspace_mem)) | |
1990 | goto out; | |
1991 | ||
1992 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1993 | if (r) |
1994 | goto out; | |
1995 | break; | |
1996 | } | |
1997 | case KVM_GET_DIRTY_LOG: { | |
1998 | struct kvm_dirty_log log; | |
1999 | ||
2000 | r = -EFAULT; | |
2f366987 | 2001 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2002 | goto out; |
2c6f5df9 | 2003 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2004 | if (r) |
2005 | goto out; | |
2006 | break; | |
2007 | } | |
5f94c174 LV |
2008 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2009 | case KVM_REGISTER_COALESCED_MMIO: { | |
2010 | struct kvm_coalesced_mmio_zone zone; | |
2011 | r = -EFAULT; | |
2012 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2013 | goto out; | |
5f94c174 LV |
2014 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
2015 | if (r) | |
2016 | goto out; | |
2017 | r = 0; | |
2018 | break; | |
2019 | } | |
2020 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
2021 | struct kvm_coalesced_mmio_zone zone; | |
2022 | r = -EFAULT; | |
2023 | if (copy_from_user(&zone, argp, sizeof zone)) | |
2024 | goto out; | |
5f94c174 LV |
2025 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
2026 | if (r) | |
2027 | goto out; | |
2028 | r = 0; | |
2029 | break; | |
2030 | } | |
2031 | #endif | |
721eecbf GH |
2032 | case KVM_IRQFD: { |
2033 | struct kvm_irqfd data; | |
2034 | ||
2035 | r = -EFAULT; | |
2036 | if (copy_from_user(&data, argp, sizeof data)) | |
2037 | goto out; | |
2038 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
2039 | break; | |
2040 | } | |
d34e6b17 GH |
2041 | case KVM_IOEVENTFD: { |
2042 | struct kvm_ioeventfd data; | |
2043 | ||
2044 | r = -EFAULT; | |
2045 | if (copy_from_user(&data, argp, sizeof data)) | |
2046 | goto out; | |
2047 | r = kvm_ioeventfd(kvm, &data); | |
2048 | break; | |
2049 | } | |
73880c80 GN |
2050 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2051 | case KVM_SET_BOOT_CPU_ID: | |
2052 | r = 0; | |
894a9c55 | 2053 | mutex_lock(&kvm->lock); |
73880c80 GN |
2054 | if (atomic_read(&kvm->online_vcpus) != 0) |
2055 | r = -EBUSY; | |
2056 | else | |
2057 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 2058 | mutex_unlock(&kvm->lock); |
73880c80 GN |
2059 | break; |
2060 | #endif | |
f17abe9a | 2061 | default: |
1fe779f8 | 2062 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
2063 | if (r == -ENOTTY) |
2064 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
2065 | } |
2066 | out: | |
2067 | return r; | |
2068 | } | |
2069 | ||
6ff5894c AB |
2070 | #ifdef CONFIG_COMPAT |
2071 | struct compat_kvm_dirty_log { | |
2072 | __u32 slot; | |
2073 | __u32 padding1; | |
2074 | union { | |
2075 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
2076 | __u64 padding2; | |
2077 | }; | |
2078 | }; | |
2079 | ||
2080 | static long kvm_vm_compat_ioctl(struct file *filp, | |
2081 | unsigned int ioctl, unsigned long arg) | |
2082 | { | |
2083 | struct kvm *kvm = filp->private_data; | |
2084 | int r; | |
2085 | ||
2086 | if (kvm->mm != current->mm) | |
2087 | return -EIO; | |
2088 | switch (ioctl) { | |
2089 | case KVM_GET_DIRTY_LOG: { | |
2090 | struct compat_kvm_dirty_log compat_log; | |
2091 | struct kvm_dirty_log log; | |
2092 | ||
2093 | r = -EFAULT; | |
2094 | if (copy_from_user(&compat_log, (void __user *)arg, | |
2095 | sizeof(compat_log))) | |
2096 | goto out; | |
2097 | log.slot = compat_log.slot; | |
2098 | log.padding1 = compat_log.padding1; | |
2099 | log.padding2 = compat_log.padding2; | |
2100 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
2101 | ||
2102 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
2103 | if (r) | |
2104 | goto out; | |
2105 | break; | |
2106 | } | |
2107 | default: | |
2108 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
2109 | } | |
2110 | ||
2111 | out: | |
2112 | return r; | |
2113 | } | |
2114 | #endif | |
2115 | ||
e4a533a4 | 2116 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 2117 | { |
777b3f49 MT |
2118 | struct page *page[1]; |
2119 | unsigned long addr; | |
2120 | int npages; | |
2121 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 2122 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 2123 | |
777b3f49 MT |
2124 | addr = gfn_to_hva(kvm, gfn); |
2125 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 2126 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2127 | |
2128 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
2129 | NULL); | |
2130 | if (unlikely(npages != 1)) | |
e4a533a4 | 2131 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2132 | |
2133 | vmf->page = page[0]; | |
e4a533a4 | 2134 | return 0; |
f17abe9a AK |
2135 | } |
2136 | ||
f0f37e2f | 2137 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 2138 | .fault = kvm_vm_fault, |
f17abe9a AK |
2139 | }; |
2140 | ||
2141 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2142 | { | |
2143 | vma->vm_ops = &kvm_vm_vm_ops; | |
2144 | return 0; | |
2145 | } | |
2146 | ||
3d3aab1b | 2147 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
2148 | .release = kvm_vm_release, |
2149 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
2150 | #ifdef CONFIG_COMPAT |
2151 | .compat_ioctl = kvm_vm_compat_ioctl, | |
2152 | #endif | |
f17abe9a | 2153 | .mmap = kvm_vm_mmap, |
6038f373 | 2154 | .llseek = noop_llseek, |
f17abe9a AK |
2155 | }; |
2156 | ||
2157 | static int kvm_dev_ioctl_create_vm(void) | |
2158 | { | |
aac87636 | 2159 | int r; |
f17abe9a AK |
2160 | struct kvm *kvm; |
2161 | ||
f17abe9a | 2162 | kvm = kvm_create_vm(); |
d6d28168 AK |
2163 | if (IS_ERR(kvm)) |
2164 | return PTR_ERR(kvm); | |
6ce5a090 TY |
2165 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2166 | r = kvm_coalesced_mmio_init(kvm); | |
2167 | if (r < 0) { | |
2168 | kvm_put_kvm(kvm); | |
2169 | return r; | |
2170 | } | |
2171 | #endif | |
aac87636 HC |
2172 | r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2173 | if (r < 0) | |
66c0b394 | 2174 | kvm_put_kvm(kvm); |
f17abe9a | 2175 | |
aac87636 | 2176 | return r; |
f17abe9a AK |
2177 | } |
2178 | ||
1a811b61 AK |
2179 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2180 | { | |
2181 | switch (arg) { | |
ca9edaee | 2182 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2183 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2184 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
2185 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2186 | case KVM_CAP_SET_BOOT_CPU_ID: | |
2187 | #endif | |
a9c7399d | 2188 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 2189 | return 1; |
399ec807 AK |
2190 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
2191 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 2192 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2193 | #endif |
1a811b61 AK |
2194 | default: |
2195 | break; | |
2196 | } | |
2197 | return kvm_dev_ioctl_check_extension(arg); | |
2198 | } | |
2199 | ||
f17abe9a AK |
2200 | static long kvm_dev_ioctl(struct file *filp, |
2201 | unsigned int ioctl, unsigned long arg) | |
2202 | { | |
07c45a36 | 2203 | long r = -EINVAL; |
f17abe9a AK |
2204 | |
2205 | switch (ioctl) { | |
2206 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2207 | r = -EINVAL; |
2208 | if (arg) | |
2209 | goto out; | |
f17abe9a AK |
2210 | r = KVM_API_VERSION; |
2211 | break; | |
2212 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2213 | r = -EINVAL; |
2214 | if (arg) | |
2215 | goto out; | |
f17abe9a AK |
2216 | r = kvm_dev_ioctl_create_vm(); |
2217 | break; | |
018d00d2 | 2218 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2219 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2220 | break; |
07c45a36 AK |
2221 | case KVM_GET_VCPU_MMAP_SIZE: |
2222 | r = -EINVAL; | |
2223 | if (arg) | |
2224 | goto out; | |
adb1ff46 AK |
2225 | r = PAGE_SIZE; /* struct kvm_run */ |
2226 | #ifdef CONFIG_X86 | |
2227 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2228 | #endif |
2229 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2230 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2231 | #endif |
07c45a36 | 2232 | break; |
d4c9ff2d FEL |
2233 | case KVM_TRACE_ENABLE: |
2234 | case KVM_TRACE_PAUSE: | |
2235 | case KVM_TRACE_DISABLE: | |
2023a29c | 2236 | r = -EOPNOTSUPP; |
d4c9ff2d | 2237 | break; |
6aa8b732 | 2238 | default: |
043405e1 | 2239 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2240 | } |
2241 | out: | |
2242 | return r; | |
2243 | } | |
2244 | ||
6aa8b732 | 2245 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2246 | .unlocked_ioctl = kvm_dev_ioctl, |
2247 | .compat_ioctl = kvm_dev_ioctl, | |
6038f373 | 2248 | .llseek = noop_llseek, |
6aa8b732 AK |
2249 | }; |
2250 | ||
2251 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2252 | KVM_MINOR, |
6aa8b732 AK |
2253 | "kvm", |
2254 | &kvm_chardev_ops, | |
2255 | }; | |
2256 | ||
75b7127c | 2257 | static void hardware_enable_nolock(void *junk) |
1b6c0168 AK |
2258 | { |
2259 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 2260 | int r; |
1b6c0168 | 2261 | |
7f59f492 | 2262 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2263 | return; |
10474ae8 | 2264 | |
7f59f492 | 2265 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
2266 | |
2267 | r = kvm_arch_hardware_enable(NULL); | |
2268 | ||
2269 | if (r) { | |
2270 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
2271 | atomic_inc(&hardware_enable_failed); | |
2272 | printk(KERN_INFO "kvm: enabling virtualization on " | |
2273 | "CPU%d failed\n", cpu); | |
2274 | } | |
1b6c0168 AK |
2275 | } |
2276 | ||
75b7127c TY |
2277 | static void hardware_enable(void *junk) |
2278 | { | |
e935b837 | 2279 | raw_spin_lock(&kvm_lock); |
75b7127c | 2280 | hardware_enable_nolock(junk); |
e935b837 | 2281 | raw_spin_unlock(&kvm_lock); |
75b7127c TY |
2282 | } |
2283 | ||
2284 | static void hardware_disable_nolock(void *junk) | |
1b6c0168 AK |
2285 | { |
2286 | int cpu = raw_smp_processor_id(); | |
2287 | ||
7f59f492 | 2288 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2289 | return; |
7f59f492 | 2290 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2291 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2292 | } |
2293 | ||
75b7127c TY |
2294 | static void hardware_disable(void *junk) |
2295 | { | |
e935b837 | 2296 | raw_spin_lock(&kvm_lock); |
75b7127c | 2297 | hardware_disable_nolock(junk); |
e935b837 | 2298 | raw_spin_unlock(&kvm_lock); |
75b7127c TY |
2299 | } |
2300 | ||
10474ae8 AG |
2301 | static void hardware_disable_all_nolock(void) |
2302 | { | |
2303 | BUG_ON(!kvm_usage_count); | |
2304 | ||
2305 | kvm_usage_count--; | |
2306 | if (!kvm_usage_count) | |
75b7127c | 2307 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
10474ae8 AG |
2308 | } |
2309 | ||
2310 | static void hardware_disable_all(void) | |
2311 | { | |
e935b837 | 2312 | raw_spin_lock(&kvm_lock); |
10474ae8 | 2313 | hardware_disable_all_nolock(); |
e935b837 | 2314 | raw_spin_unlock(&kvm_lock); |
10474ae8 AG |
2315 | } |
2316 | ||
2317 | static int hardware_enable_all(void) | |
2318 | { | |
2319 | int r = 0; | |
2320 | ||
e935b837 | 2321 | raw_spin_lock(&kvm_lock); |
10474ae8 AG |
2322 | |
2323 | kvm_usage_count++; | |
2324 | if (kvm_usage_count == 1) { | |
2325 | atomic_set(&hardware_enable_failed, 0); | |
75b7127c | 2326 | on_each_cpu(hardware_enable_nolock, NULL, 1); |
10474ae8 AG |
2327 | |
2328 | if (atomic_read(&hardware_enable_failed)) { | |
2329 | hardware_disable_all_nolock(); | |
2330 | r = -EBUSY; | |
2331 | } | |
2332 | } | |
2333 | ||
e935b837 | 2334 | raw_spin_unlock(&kvm_lock); |
10474ae8 AG |
2335 | |
2336 | return r; | |
2337 | } | |
2338 | ||
774c47f1 AK |
2339 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2340 | void *v) | |
2341 | { | |
2342 | int cpu = (long)v; | |
2343 | ||
10474ae8 AG |
2344 | if (!kvm_usage_count) |
2345 | return NOTIFY_OK; | |
2346 | ||
1a6f4d7f | 2347 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2348 | switch (val) { |
cec9ad27 | 2349 | case CPU_DYING: |
6ec8a856 AK |
2350 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2351 | cpu); | |
2352 | hardware_disable(NULL); | |
2353 | break; | |
da908f2f | 2354 | case CPU_STARTING: |
43934a38 JK |
2355 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2356 | cpu); | |
da908f2f | 2357 | hardware_enable(NULL); |
774c47f1 AK |
2358 | break; |
2359 | } | |
2360 | return NOTIFY_OK; | |
2361 | } | |
2362 | ||
4ecac3fd | 2363 | |
b7c4145b | 2364 | asmlinkage void kvm_spurious_fault(void) |
4ecac3fd | 2365 | { |
4ecac3fd AK |
2366 | /* Fault while not rebooting. We want the trace. */ |
2367 | BUG(); | |
2368 | } | |
b7c4145b | 2369 | EXPORT_SYMBOL_GPL(kvm_spurious_fault); |
4ecac3fd | 2370 | |
9a2b85c6 | 2371 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2372 | void *v) |
9a2b85c6 | 2373 | { |
8e1c1815 SY |
2374 | /* |
2375 | * Some (well, at least mine) BIOSes hang on reboot if | |
2376 | * in vmx root mode. | |
2377 | * | |
2378 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2379 | */ | |
2380 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2381 | kvm_rebooting = true; | |
75b7127c | 2382 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
9a2b85c6 RR |
2383 | return NOTIFY_OK; |
2384 | } | |
2385 | ||
2386 | static struct notifier_block kvm_reboot_notifier = { | |
2387 | .notifier_call = kvm_reboot, | |
2388 | .priority = 0, | |
2389 | }; | |
2390 | ||
e93f8a0f | 2391 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
2392 | { |
2393 | int i; | |
2394 | ||
2395 | for (i = 0; i < bus->dev_count; i++) { | |
743eeb0b | 2396 | struct kvm_io_device *pos = bus->range[i].dev; |
2eeb2e94 GH |
2397 | |
2398 | kvm_iodevice_destructor(pos); | |
2399 | } | |
e93f8a0f | 2400 | kfree(bus); |
2eeb2e94 GH |
2401 | } |
2402 | ||
743eeb0b SL |
2403 | int kvm_io_bus_sort_cmp(const void *p1, const void *p2) |
2404 | { | |
2405 | const struct kvm_io_range *r1 = p1; | |
2406 | const struct kvm_io_range *r2 = p2; | |
2407 | ||
2408 | if (r1->addr < r2->addr) | |
2409 | return -1; | |
2410 | if (r1->addr + r1->len > r2->addr + r2->len) | |
2411 | return 1; | |
2412 | return 0; | |
2413 | } | |
2414 | ||
2415 | int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev, | |
2416 | gpa_t addr, int len) | |
2417 | { | |
2418 | if (bus->dev_count == NR_IOBUS_DEVS) | |
2419 | return -ENOSPC; | |
2420 | ||
2421 | bus->range[bus->dev_count++] = (struct kvm_io_range) { | |
2422 | .addr = addr, | |
2423 | .len = len, | |
2424 | .dev = dev, | |
2425 | }; | |
2426 | ||
2427 | sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range), | |
2428 | kvm_io_bus_sort_cmp, NULL); | |
2429 | ||
2430 | return 0; | |
2431 | } | |
2432 | ||
2433 | int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus, | |
2434 | gpa_t addr, int len) | |
2435 | { | |
2436 | struct kvm_io_range *range, key; | |
2437 | int off; | |
2438 | ||
2439 | key = (struct kvm_io_range) { | |
2440 | .addr = addr, | |
2441 | .len = len, | |
2442 | }; | |
2443 | ||
2444 | range = bsearch(&key, bus->range, bus->dev_count, | |
2445 | sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp); | |
2446 | if (range == NULL) | |
2447 | return -ENOENT; | |
2448 | ||
2449 | off = range - bus->range; | |
2450 | ||
2451 | while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0) | |
2452 | off--; | |
2453 | ||
2454 | return off; | |
2455 | } | |
2456 | ||
bda9020e | 2457 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2458 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2459 | int len, const void *val) |
2eeb2e94 | 2460 | { |
743eeb0b | 2461 | int idx; |
90d83dc3 | 2462 | struct kvm_io_bus *bus; |
743eeb0b SL |
2463 | struct kvm_io_range range; |
2464 | ||
2465 | range = (struct kvm_io_range) { | |
2466 | .addr = addr, | |
2467 | .len = len, | |
2468 | }; | |
90d83dc3 LJ |
2469 | |
2470 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
743eeb0b SL |
2471 | idx = kvm_io_bus_get_first_dev(bus, addr, len); |
2472 | if (idx < 0) | |
2473 | return -EOPNOTSUPP; | |
2474 | ||
2475 | while (idx < bus->dev_count && | |
2476 | kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) { | |
2477 | if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val)) | |
bda9020e | 2478 | return 0; |
743eeb0b SL |
2479 | idx++; |
2480 | } | |
2481 | ||
bda9020e MT |
2482 | return -EOPNOTSUPP; |
2483 | } | |
2eeb2e94 | 2484 | |
bda9020e | 2485 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2486 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2487 | int len, void *val) | |
bda9020e | 2488 | { |
743eeb0b | 2489 | int idx; |
90d83dc3 | 2490 | struct kvm_io_bus *bus; |
743eeb0b SL |
2491 | struct kvm_io_range range; |
2492 | ||
2493 | range = (struct kvm_io_range) { | |
2494 | .addr = addr, | |
2495 | .len = len, | |
2496 | }; | |
e93f8a0f | 2497 | |
90d83dc3 | 2498 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
743eeb0b SL |
2499 | idx = kvm_io_bus_get_first_dev(bus, addr, len); |
2500 | if (idx < 0) | |
2501 | return -EOPNOTSUPP; | |
2502 | ||
2503 | while (idx < bus->dev_count && | |
2504 | kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) { | |
2505 | if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val)) | |
bda9020e | 2506 | return 0; |
743eeb0b SL |
2507 | idx++; |
2508 | } | |
2509 | ||
bda9020e | 2510 | return -EOPNOTSUPP; |
2eeb2e94 GH |
2511 | } |
2512 | ||
79fac95e | 2513 | /* Caller must hold slots_lock. */ |
743eeb0b SL |
2514 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2515 | int len, struct kvm_io_device *dev) | |
6c474694 | 2516 | { |
e93f8a0f | 2517 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2518 | |
e93f8a0f | 2519 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2520 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2521 | return -ENOSPC; | |
2eeb2e94 | 2522 | |
e93f8a0f MT |
2523 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2524 | if (!new_bus) | |
2525 | return -ENOMEM; | |
2526 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
743eeb0b | 2527 | kvm_io_bus_insert_dev(new_bus, dev, addr, len); |
e93f8a0f MT |
2528 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); |
2529 | synchronize_srcu_expedited(&kvm->srcu); | |
2530 | kfree(bus); | |
090b7aff GH |
2531 | |
2532 | return 0; | |
2533 | } | |
2534 | ||
79fac95e | 2535 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2536 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2537 | struct kvm_io_device *dev) | |
090b7aff | 2538 | { |
e93f8a0f MT |
2539 | int i, r; |
2540 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2541 | |
e93f8a0f MT |
2542 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2543 | if (!new_bus) | |
2544 | return -ENOMEM; | |
090b7aff | 2545 | |
e93f8a0f MT |
2546 | bus = kvm->buses[bus_idx]; |
2547 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2548 | ||
2549 | r = -ENOENT; | |
2550 | for (i = 0; i < new_bus->dev_count; i++) | |
743eeb0b | 2551 | if (new_bus->range[i].dev == dev) { |
e93f8a0f | 2552 | r = 0; |
743eeb0b SL |
2553 | new_bus->dev_count--; |
2554 | new_bus->range[i] = new_bus->range[new_bus->dev_count]; | |
2555 | sort(new_bus->range, new_bus->dev_count, | |
2556 | sizeof(struct kvm_io_range), | |
2557 | kvm_io_bus_sort_cmp, NULL); | |
090b7aff GH |
2558 | break; |
2559 | } | |
e93f8a0f MT |
2560 | |
2561 | if (r) { | |
2562 | kfree(new_bus); | |
2563 | return r; | |
2564 | } | |
2565 | ||
2566 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2567 | synchronize_srcu_expedited(&kvm->srcu); | |
2568 | kfree(bus); | |
2569 | return r; | |
2eeb2e94 GH |
2570 | } |
2571 | ||
774c47f1 AK |
2572 | static struct notifier_block kvm_cpu_notifier = { |
2573 | .notifier_call = kvm_cpu_hotplug, | |
774c47f1 AK |
2574 | }; |
2575 | ||
8b88b099 | 2576 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2577 | { |
2578 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2579 | struct kvm *kvm; |
2580 | ||
8b88b099 | 2581 | *val = 0; |
e935b837 | 2582 | raw_spin_lock(&kvm_lock); |
ba1389b7 | 2583 | list_for_each_entry(kvm, &vm_list, vm_list) |
8b88b099 | 2584 | *val += *(u32 *)((void *)kvm + offset); |
e935b837 | 2585 | raw_spin_unlock(&kvm_lock); |
8b88b099 | 2586 | return 0; |
ba1389b7 AK |
2587 | } |
2588 | ||
2589 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2590 | ||
8b88b099 | 2591 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2592 | { |
2593 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2594 | struct kvm *kvm; |
2595 | struct kvm_vcpu *vcpu; | |
2596 | int i; | |
2597 | ||
8b88b099 | 2598 | *val = 0; |
e935b837 | 2599 | raw_spin_lock(&kvm_lock); |
1165f5fe | 2600 | list_for_each_entry(kvm, &vm_list, vm_list) |
988a2cae GN |
2601 | kvm_for_each_vcpu(i, vcpu, kvm) |
2602 | *val += *(u32 *)((void *)vcpu + offset); | |
2603 | ||
e935b837 | 2604 | raw_spin_unlock(&kvm_lock); |
8b88b099 | 2605 | return 0; |
1165f5fe AK |
2606 | } |
2607 | ||
ba1389b7 AK |
2608 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2609 | ||
828c0950 | 2610 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2611 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2612 | [KVM_STAT_VM] = &vm_stat_fops, | |
2613 | }; | |
1165f5fe | 2614 | |
a16b043c | 2615 | static void kvm_init_debug(void) |
6aa8b732 AK |
2616 | { |
2617 | struct kvm_stats_debugfs_item *p; | |
2618 | ||
76f7c879 | 2619 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2620 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2621 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2622 | (void *)(long)p->offset, |
ba1389b7 | 2623 | stat_fops[p->kind]); |
6aa8b732 AK |
2624 | } |
2625 | ||
2626 | static void kvm_exit_debug(void) | |
2627 | { | |
2628 | struct kvm_stats_debugfs_item *p; | |
2629 | ||
2630 | for (p = debugfs_entries; p->name; ++p) | |
2631 | debugfs_remove(p->dentry); | |
76f7c879 | 2632 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2633 | } |
2634 | ||
fb3600cc | 2635 | static int kvm_suspend(void) |
59ae6c6b | 2636 | { |
10474ae8 | 2637 | if (kvm_usage_count) |
75b7127c | 2638 | hardware_disable_nolock(NULL); |
59ae6c6b AK |
2639 | return 0; |
2640 | } | |
2641 | ||
fb3600cc | 2642 | static void kvm_resume(void) |
59ae6c6b | 2643 | { |
ca84d1a2 | 2644 | if (kvm_usage_count) { |
e935b837 | 2645 | WARN_ON(raw_spin_is_locked(&kvm_lock)); |
75b7127c | 2646 | hardware_enable_nolock(NULL); |
ca84d1a2 | 2647 | } |
59ae6c6b AK |
2648 | } |
2649 | ||
fb3600cc | 2650 | static struct syscore_ops kvm_syscore_ops = { |
59ae6c6b AK |
2651 | .suspend = kvm_suspend, |
2652 | .resume = kvm_resume, | |
2653 | }; | |
2654 | ||
cea7bb21 | 2655 | struct page *bad_page; |
35149e21 | 2656 | pfn_t bad_pfn; |
6aa8b732 | 2657 | |
15ad7146 AK |
2658 | static inline |
2659 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2660 | { | |
2661 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2662 | } | |
2663 | ||
2664 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2665 | { | |
2666 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2667 | ||
e9b11c17 | 2668 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2669 | } |
2670 | ||
2671 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2672 | struct task_struct *next) | |
2673 | { | |
2674 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2675 | ||
e9b11c17 | 2676 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2677 | } |
2678 | ||
0ee75bea | 2679 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2680 | struct module *module) |
6aa8b732 AK |
2681 | { |
2682 | int r; | |
002c7f7c | 2683 | int cpu; |
6aa8b732 | 2684 | |
f8c16bba ZX |
2685 | r = kvm_arch_init(opaque); |
2686 | if (r) | |
d2308784 | 2687 | goto out_fail; |
cb498ea2 ZX |
2688 | |
2689 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2690 | ||
2691 | if (bad_page == NULL) { | |
2692 | r = -ENOMEM; | |
2693 | goto out; | |
2694 | } | |
2695 | ||
35149e21 AL |
2696 | bad_pfn = page_to_pfn(bad_page); |
2697 | ||
bf998156 HY |
2698 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2699 | ||
2700 | if (hwpoison_page == NULL) { | |
2701 | r = -ENOMEM; | |
2702 | goto out_free_0; | |
2703 | } | |
2704 | ||
2705 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2706 | ||
edba23e5 GN |
2707 | fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2708 | ||
2709 | if (fault_page == NULL) { | |
2710 | r = -ENOMEM; | |
2711 | goto out_free_0; | |
2712 | } | |
2713 | ||
2714 | fault_pfn = page_to_pfn(fault_page); | |
2715 | ||
8437a617 | 2716 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2717 | r = -ENOMEM; |
2718 | goto out_free_0; | |
2719 | } | |
2720 | ||
e9b11c17 | 2721 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2722 | if (r < 0) |
7f59f492 | 2723 | goto out_free_0a; |
6aa8b732 | 2724 | |
002c7f7c YS |
2725 | for_each_online_cpu(cpu) { |
2726 | smp_call_function_single(cpu, | |
e9b11c17 | 2727 | kvm_arch_check_processor_compat, |
8691e5a8 | 2728 | &r, 1); |
002c7f7c | 2729 | if (r < 0) |
d2308784 | 2730 | goto out_free_1; |
002c7f7c YS |
2731 | } |
2732 | ||
774c47f1 AK |
2733 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2734 | if (r) | |
d2308784 | 2735 | goto out_free_2; |
6aa8b732 AK |
2736 | register_reboot_notifier(&kvm_reboot_notifier); |
2737 | ||
c16f862d | 2738 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2739 | if (!vcpu_align) |
2740 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2741 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2742 | 0, NULL); |
c16f862d RR |
2743 | if (!kvm_vcpu_cache) { |
2744 | r = -ENOMEM; | |
fb3600cc | 2745 | goto out_free_3; |
c16f862d RR |
2746 | } |
2747 | ||
af585b92 GN |
2748 | r = kvm_async_pf_init(); |
2749 | if (r) | |
2750 | goto out_free; | |
2751 | ||
6aa8b732 | 2752 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2753 | kvm_vm_fops.owner = module; |
2754 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2755 | |
2756 | r = misc_register(&kvm_dev); | |
2757 | if (r) { | |
d77c26fc | 2758 | printk(KERN_ERR "kvm: misc device register failed\n"); |
af585b92 | 2759 | goto out_unreg; |
6aa8b732 AK |
2760 | } |
2761 | ||
fb3600cc RW |
2762 | register_syscore_ops(&kvm_syscore_ops); |
2763 | ||
15ad7146 AK |
2764 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2765 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2766 | ||
0ea4ed8e DW |
2767 | kvm_init_debug(); |
2768 | ||
c7addb90 | 2769 | return 0; |
6aa8b732 | 2770 | |
af585b92 GN |
2771 | out_unreg: |
2772 | kvm_async_pf_deinit(); | |
6aa8b732 | 2773 | out_free: |
c16f862d | 2774 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2775 | out_free_3: |
6aa8b732 | 2776 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2777 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2778 | out_free_2: |
d2308784 | 2779 | out_free_1: |
e9b11c17 | 2780 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2781 | out_free_0a: |
2782 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2783 | out_free_0: |
edba23e5 GN |
2784 | if (fault_page) |
2785 | __free_page(fault_page); | |
bf998156 HY |
2786 | if (hwpoison_page) |
2787 | __free_page(hwpoison_page); | |
d2308784 | 2788 | __free_page(bad_page); |
ca45aaae | 2789 | out: |
f8c16bba | 2790 | kvm_arch_exit(); |
d2308784 | 2791 | out_fail: |
6aa8b732 AK |
2792 | return r; |
2793 | } | |
cb498ea2 | 2794 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2795 | |
cb498ea2 | 2796 | void kvm_exit(void) |
6aa8b732 | 2797 | { |
0ea4ed8e | 2798 | kvm_exit_debug(); |
6aa8b732 | 2799 | misc_deregister(&kvm_dev); |
c16f862d | 2800 | kmem_cache_destroy(kvm_vcpu_cache); |
af585b92 | 2801 | kvm_async_pf_deinit(); |
fb3600cc | 2802 | unregister_syscore_ops(&kvm_syscore_ops); |
6aa8b732 | 2803 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2804 | unregister_cpu_notifier(&kvm_cpu_notifier); |
75b7127c | 2805 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
e9b11c17 | 2806 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2807 | kvm_arch_exit(); |
7f59f492 | 2808 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2809 | __free_page(hwpoison_page); |
cea7bb21 | 2810 | __free_page(bad_page); |
6aa8b732 | 2811 | } |
cb498ea2 | 2812 | EXPORT_SYMBOL_GPL(kvm_exit); |