Commit | Line | Data |
---|---|---|
6aa8b732 AK |
1 | /* |
2 | * Kernel-based Virtual Machine driver for Linux | |
3 | * | |
4 | * This module enables machines with Intel VT-x extensions to run virtual | |
5 | * machines without emulation or binary translation. | |
6 | * | |
7 | * Copyright (C) 2006 Qumranet, Inc. | |
9611c187 | 8 | * Copyright 2010 Red Hat, Inc. and/or its affiliates. |
6aa8b732 AK |
9 | * |
10 | * Authors: | |
11 | * Avi Kivity <avi@qumranet.com> | |
12 | * Yaniv Kamay <yaniv@qumranet.com> | |
13 | * | |
14 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
15 | * the COPYING file in the top-level directory. | |
16 | * | |
17 | */ | |
18 | ||
e2174021 | 19 | #include "iodev.h" |
6aa8b732 | 20 | |
edf88417 | 21 | #include <linux/kvm_host.h> |
6aa8b732 AK |
22 | #include <linux/kvm.h> |
23 | #include <linux/module.h> | |
24 | #include <linux/errno.h> | |
6aa8b732 | 25 | #include <linux/percpu.h> |
6aa8b732 AK |
26 | #include <linux/mm.h> |
27 | #include <linux/miscdevice.h> | |
28 | #include <linux/vmalloc.h> | |
6aa8b732 | 29 | #include <linux/reboot.h> |
6aa8b732 AK |
30 | #include <linux/debugfs.h> |
31 | #include <linux/highmem.h> | |
32 | #include <linux/file.h> | |
59ae6c6b | 33 | #include <linux/sysdev.h> |
774c47f1 | 34 | #include <linux/cpu.h> |
e8edc6e0 | 35 | #include <linux/sched.h> |
d9e368d6 AK |
36 | #include <linux/cpumask.h> |
37 | #include <linux/smp.h> | |
d6d28168 | 38 | #include <linux/anon_inodes.h> |
04d2cc77 | 39 | #include <linux/profile.h> |
7aa81cc0 | 40 | #include <linux/kvm_para.h> |
6fc138d2 | 41 | #include <linux/pagemap.h> |
8d4e1288 | 42 | #include <linux/mman.h> |
35149e21 | 43 | #include <linux/swap.h> |
e56d532f | 44 | #include <linux/bitops.h> |
547de29e | 45 | #include <linux/spinlock.h> |
6ff5894c | 46 | #include <linux/compat.h> |
bc6678a3 | 47 | #include <linux/srcu.h> |
8f0b1ab6 | 48 | #include <linux/hugetlb.h> |
5a0e3ad6 | 49 | #include <linux/slab.h> |
6aa8b732 | 50 | |
e495606d | 51 | #include <asm/processor.h> |
e495606d AK |
52 | #include <asm/io.h> |
53 | #include <asm/uaccess.h> | |
3e021bf5 | 54 | #include <asm/pgtable.h> |
c8240bd6 | 55 | #include <asm-generic/bitops/le.h> |
6aa8b732 | 56 | |
5f94c174 | 57 | #include "coalesced_mmio.h" |
af585b92 | 58 | #include "async_pf.h" |
5f94c174 | 59 | |
229456fc MT |
60 | #define CREATE_TRACE_POINTS |
61 | #include <trace/events/kvm.h> | |
62 | ||
6aa8b732 AK |
63 | MODULE_AUTHOR("Qumranet"); |
64 | MODULE_LICENSE("GPL"); | |
65 | ||
fa40a821 MT |
66 | /* |
67 | * Ordering of locks: | |
68 | * | |
fae3a353 | 69 | * kvm->lock --> kvm->slots_lock --> kvm->irq_lock |
fa40a821 MT |
70 | */ |
71 | ||
e9b11c17 ZX |
72 | DEFINE_SPINLOCK(kvm_lock); |
73 | LIST_HEAD(vm_list); | |
133de902 | 74 | |
7f59f492 | 75 | static cpumask_var_t cpus_hardware_enabled; |
10474ae8 AG |
76 | static int kvm_usage_count = 0; |
77 | static atomic_t hardware_enable_failed; | |
1b6c0168 | 78 | |
c16f862d RR |
79 | struct kmem_cache *kvm_vcpu_cache; |
80 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 81 | |
15ad7146 AK |
82 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
83 | ||
76f7c879 | 84 | struct dentry *kvm_debugfs_dir; |
6aa8b732 | 85 | |
bccf2150 AK |
86 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
87 | unsigned long arg); | |
10474ae8 AG |
88 | static int hardware_enable_all(void); |
89 | static void hardware_disable_all(void); | |
bccf2150 | 90 | |
e93f8a0f MT |
91 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus); |
92 | ||
b7c4145b AK |
93 | bool kvm_rebooting; |
94 | EXPORT_SYMBOL_GPL(kvm_rebooting); | |
4ecac3fd | 95 | |
54dee993 MT |
96 | static bool largepages_enabled = true; |
97 | ||
fa7bff8f GN |
98 | static struct page *hwpoison_page; |
99 | static pfn_t hwpoison_pfn; | |
bf998156 | 100 | |
edba23e5 GN |
101 | static struct page *fault_page; |
102 | static pfn_t fault_pfn; | |
103 | ||
c77fb9dc | 104 | inline int kvm_is_mmio_pfn(pfn_t pfn) |
cbff90a7 | 105 | { |
fc5659c8 | 106 | if (pfn_valid(pfn)) { |
22e5c47e | 107 | int reserved; |
936a5fe6 | 108 | struct page *tail = pfn_to_page(pfn); |
22e5c47e AA |
109 | struct page *head = compound_trans_head(tail); |
110 | reserved = PageReserved(head); | |
936a5fe6 | 111 | if (head != tail) { |
936a5fe6 | 112 | /* |
22e5c47e AA |
113 | * "head" is not a dangling pointer |
114 | * (compound_trans_head takes care of that) | |
115 | * but the hugepage may have been splitted | |
116 | * from under us (and we may not hold a | |
117 | * reference count on the head page so it can | |
118 | * be reused before we run PageReferenced), so | |
119 | * we've to check PageTail before returning | |
120 | * what we just read. | |
936a5fe6 | 121 | */ |
22e5c47e AA |
122 | smp_rmb(); |
123 | if (PageTail(tail)) | |
124 | return reserved; | |
936a5fe6 AA |
125 | } |
126 | return PageReserved(tail); | |
fc5659c8 | 127 | } |
cbff90a7 BAY |
128 | |
129 | return true; | |
130 | } | |
131 | ||
bccf2150 AK |
132 | /* |
133 | * Switches to specified vcpu, until a matching vcpu_put() | |
134 | */ | |
313a3dc7 | 135 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 136 | { |
15ad7146 AK |
137 | int cpu; |
138 | ||
bccf2150 | 139 | mutex_lock(&vcpu->mutex); |
34bb10b7 RR |
140 | if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) { |
141 | /* The thread running this VCPU changed. */ | |
142 | struct pid *oldpid = vcpu->pid; | |
143 | struct pid *newpid = get_task_pid(current, PIDTYPE_PID); | |
144 | rcu_assign_pointer(vcpu->pid, newpid); | |
145 | synchronize_rcu(); | |
146 | put_pid(oldpid); | |
147 | } | |
15ad7146 AK |
148 | cpu = get_cpu(); |
149 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 150 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 151 | put_cpu(); |
6aa8b732 AK |
152 | } |
153 | ||
313a3dc7 | 154 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 155 | { |
15ad7146 | 156 | preempt_disable(); |
313a3dc7 | 157 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
158 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
159 | preempt_enable(); | |
6aa8b732 AK |
160 | mutex_unlock(&vcpu->mutex); |
161 | } | |
162 | ||
d9e368d6 AK |
163 | static void ack_flush(void *_completed) |
164 | { | |
d9e368d6 AK |
165 | } |
166 | ||
49846896 | 167 | static bool make_all_cpus_request(struct kvm *kvm, unsigned int req) |
d9e368d6 | 168 | { |
597a5f55 | 169 | int i, cpu, me; |
6ef7a1bc RR |
170 | cpumask_var_t cpus; |
171 | bool called = true; | |
d9e368d6 | 172 | struct kvm_vcpu *vcpu; |
d9e368d6 | 173 | |
79f55997 | 174 | zalloc_cpumask_var(&cpus, GFP_ATOMIC); |
6ef7a1bc | 175 | |
3cba4130 | 176 | me = get_cpu(); |
988a2cae | 177 | kvm_for_each_vcpu(i, vcpu, kvm) { |
3cba4130 | 178 | kvm_make_request(req, vcpu); |
d9e368d6 | 179 | cpu = vcpu->cpu; |
6b7e2d09 XG |
180 | |
181 | /* Set ->requests bit before we read ->mode */ | |
182 | smp_mb(); | |
183 | ||
184 | if (cpus != NULL && cpu != -1 && cpu != me && | |
185 | kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE) | |
6ef7a1bc | 186 | cpumask_set_cpu(cpu, cpus); |
49846896 | 187 | } |
6ef7a1bc RR |
188 | if (unlikely(cpus == NULL)) |
189 | smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1); | |
190 | else if (!cpumask_empty(cpus)) | |
191 | smp_call_function_many(cpus, ack_flush, NULL, 1); | |
192 | else | |
193 | called = false; | |
3cba4130 | 194 | put_cpu(); |
6ef7a1bc | 195 | free_cpumask_var(cpus); |
49846896 | 196 | return called; |
d9e368d6 AK |
197 | } |
198 | ||
49846896 | 199 | void kvm_flush_remote_tlbs(struct kvm *kvm) |
2e53d63a | 200 | { |
a4ee1ca4 XG |
201 | int dirty_count = kvm->tlbs_dirty; |
202 | ||
203 | smp_mb(); | |
49846896 RR |
204 | if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH)) |
205 | ++kvm->stat.remote_tlb_flush; | |
a4ee1ca4 | 206 | cmpxchg(&kvm->tlbs_dirty, dirty_count, 0); |
2e53d63a MT |
207 | } |
208 | ||
49846896 RR |
209 | void kvm_reload_remote_mmus(struct kvm *kvm) |
210 | { | |
211 | make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD); | |
212 | } | |
2e53d63a | 213 | |
fb3f0f51 RR |
214 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
215 | { | |
216 | struct page *page; | |
217 | int r; | |
218 | ||
219 | mutex_init(&vcpu->mutex); | |
220 | vcpu->cpu = -1; | |
fb3f0f51 RR |
221 | vcpu->kvm = kvm; |
222 | vcpu->vcpu_id = id; | |
34bb10b7 | 223 | vcpu->pid = NULL; |
b6958ce4 | 224 | init_waitqueue_head(&vcpu->wq); |
af585b92 | 225 | kvm_async_pf_vcpu_init(vcpu); |
fb3f0f51 RR |
226 | |
227 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
228 | if (!page) { | |
229 | r = -ENOMEM; | |
230 | goto fail; | |
231 | } | |
232 | vcpu->run = page_address(page); | |
233 | ||
e9b11c17 | 234 | r = kvm_arch_vcpu_init(vcpu); |
fb3f0f51 | 235 | if (r < 0) |
e9b11c17 | 236 | goto fail_free_run; |
fb3f0f51 RR |
237 | return 0; |
238 | ||
fb3f0f51 RR |
239 | fail_free_run: |
240 | free_page((unsigned long)vcpu->run); | |
241 | fail: | |
76fafa5e | 242 | return r; |
fb3f0f51 RR |
243 | } |
244 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
245 | ||
246 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
247 | { | |
34bb10b7 | 248 | put_pid(vcpu->pid); |
e9b11c17 | 249 | kvm_arch_vcpu_uninit(vcpu); |
fb3f0f51 RR |
250 | free_page((unsigned long)vcpu->run); |
251 | } | |
252 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
253 | ||
e930bffe AA |
254 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
255 | static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn) | |
256 | { | |
257 | return container_of(mn, struct kvm, mmu_notifier); | |
258 | } | |
259 | ||
260 | static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn, | |
261 | struct mm_struct *mm, | |
262 | unsigned long address) | |
263 | { | |
264 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 265 | int need_tlb_flush, idx; |
e930bffe AA |
266 | |
267 | /* | |
268 | * When ->invalidate_page runs, the linux pte has been zapped | |
269 | * already but the page is still allocated until | |
270 | * ->invalidate_page returns. So if we increase the sequence | |
271 | * here the kvm page fault will notice if the spte can't be | |
272 | * established because the page is going to be freed. If | |
273 | * instead the kvm page fault establishes the spte before | |
274 | * ->invalidate_page runs, kvm_unmap_hva will release it | |
275 | * before returning. | |
276 | * | |
277 | * The sequence increase only need to be seen at spin_unlock | |
278 | * time, and not at spin_lock time. | |
279 | * | |
280 | * Increasing the sequence after the spin_unlock would be | |
281 | * unsafe because the kvm page fault could then establish the | |
282 | * pte after kvm_unmap_hva returned, without noticing the page | |
283 | * is going to be freed. | |
284 | */ | |
bc6678a3 | 285 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
286 | spin_lock(&kvm->mmu_lock); |
287 | kvm->mmu_notifier_seq++; | |
a4ee1ca4 | 288 | need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty; |
e930bffe | 289 | spin_unlock(&kvm->mmu_lock); |
bc6678a3 | 290 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
291 | |
292 | /* we've to flush the tlb before the pages can be freed */ | |
293 | if (need_tlb_flush) | |
294 | kvm_flush_remote_tlbs(kvm); | |
295 | ||
296 | } | |
297 | ||
3da0dd43 IE |
298 | static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn, |
299 | struct mm_struct *mm, | |
300 | unsigned long address, | |
301 | pte_t pte) | |
302 | { | |
303 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 304 | int idx; |
3da0dd43 | 305 | |
bc6678a3 | 306 | idx = srcu_read_lock(&kvm->srcu); |
3da0dd43 IE |
307 | spin_lock(&kvm->mmu_lock); |
308 | kvm->mmu_notifier_seq++; | |
309 | kvm_set_spte_hva(kvm, address, pte); | |
310 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 311 | srcu_read_unlock(&kvm->srcu, idx); |
3da0dd43 IE |
312 | } |
313 | ||
e930bffe AA |
314 | static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn, |
315 | struct mm_struct *mm, | |
316 | unsigned long start, | |
317 | unsigned long end) | |
318 | { | |
319 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 320 | int need_tlb_flush = 0, idx; |
e930bffe | 321 | |
bc6678a3 | 322 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
323 | spin_lock(&kvm->mmu_lock); |
324 | /* | |
325 | * The count increase must become visible at unlock time as no | |
326 | * spte can be established without taking the mmu_lock and | |
327 | * count is also read inside the mmu_lock critical section. | |
328 | */ | |
329 | kvm->mmu_notifier_count++; | |
330 | for (; start < end; start += PAGE_SIZE) | |
331 | need_tlb_flush |= kvm_unmap_hva(kvm, start); | |
a4ee1ca4 | 332 | need_tlb_flush |= kvm->tlbs_dirty; |
e930bffe | 333 | spin_unlock(&kvm->mmu_lock); |
bc6678a3 | 334 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
335 | |
336 | /* we've to flush the tlb before the pages can be freed */ | |
337 | if (need_tlb_flush) | |
338 | kvm_flush_remote_tlbs(kvm); | |
339 | } | |
340 | ||
341 | static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn, | |
342 | struct mm_struct *mm, | |
343 | unsigned long start, | |
344 | unsigned long end) | |
345 | { | |
346 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
347 | ||
348 | spin_lock(&kvm->mmu_lock); | |
349 | /* | |
350 | * This sequence increase will notify the kvm page fault that | |
351 | * the page that is going to be mapped in the spte could have | |
352 | * been freed. | |
353 | */ | |
354 | kvm->mmu_notifier_seq++; | |
355 | /* | |
356 | * The above sequence increase must be visible before the | |
357 | * below count decrease but both values are read by the kvm | |
358 | * page fault under mmu_lock spinlock so we don't need to add | |
359 | * a smb_wmb() here in between the two. | |
360 | */ | |
361 | kvm->mmu_notifier_count--; | |
362 | spin_unlock(&kvm->mmu_lock); | |
363 | ||
364 | BUG_ON(kvm->mmu_notifier_count < 0); | |
365 | } | |
366 | ||
367 | static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn, | |
368 | struct mm_struct *mm, | |
369 | unsigned long address) | |
370 | { | |
371 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
bc6678a3 | 372 | int young, idx; |
e930bffe | 373 | |
bc6678a3 | 374 | idx = srcu_read_lock(&kvm->srcu); |
e930bffe AA |
375 | spin_lock(&kvm->mmu_lock); |
376 | young = kvm_age_hva(kvm, address); | |
377 | spin_unlock(&kvm->mmu_lock); | |
bc6678a3 | 378 | srcu_read_unlock(&kvm->srcu, idx); |
e930bffe AA |
379 | |
380 | if (young) | |
381 | kvm_flush_remote_tlbs(kvm); | |
382 | ||
383 | return young; | |
384 | } | |
385 | ||
8ee53820 AA |
386 | static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn, |
387 | struct mm_struct *mm, | |
388 | unsigned long address) | |
389 | { | |
390 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
391 | int young, idx; | |
392 | ||
393 | idx = srcu_read_lock(&kvm->srcu); | |
394 | spin_lock(&kvm->mmu_lock); | |
395 | young = kvm_test_age_hva(kvm, address); | |
396 | spin_unlock(&kvm->mmu_lock); | |
397 | srcu_read_unlock(&kvm->srcu, idx); | |
398 | ||
399 | return young; | |
400 | } | |
401 | ||
85db06e5 MT |
402 | static void kvm_mmu_notifier_release(struct mmu_notifier *mn, |
403 | struct mm_struct *mm) | |
404 | { | |
405 | struct kvm *kvm = mmu_notifier_to_kvm(mn); | |
eda2beda LJ |
406 | int idx; |
407 | ||
408 | idx = srcu_read_lock(&kvm->srcu); | |
85db06e5 | 409 | kvm_arch_flush_shadow(kvm); |
eda2beda | 410 | srcu_read_unlock(&kvm->srcu, idx); |
85db06e5 MT |
411 | } |
412 | ||
e930bffe AA |
413 | static const struct mmu_notifier_ops kvm_mmu_notifier_ops = { |
414 | .invalidate_page = kvm_mmu_notifier_invalidate_page, | |
415 | .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start, | |
416 | .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end, | |
417 | .clear_flush_young = kvm_mmu_notifier_clear_flush_young, | |
8ee53820 | 418 | .test_young = kvm_mmu_notifier_test_young, |
3da0dd43 | 419 | .change_pte = kvm_mmu_notifier_change_pte, |
85db06e5 | 420 | .release = kvm_mmu_notifier_release, |
e930bffe | 421 | }; |
4c07b0a4 AK |
422 | |
423 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
424 | { | |
425 | kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops; | |
426 | return mmu_notifier_register(&kvm->mmu_notifier, current->mm); | |
427 | } | |
428 | ||
429 | #else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */ | |
430 | ||
431 | static int kvm_init_mmu_notifier(struct kvm *kvm) | |
432 | { | |
433 | return 0; | |
434 | } | |
435 | ||
e930bffe AA |
436 | #endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */ |
437 | ||
f17abe9a | 438 | static struct kvm *kvm_create_vm(void) |
6aa8b732 | 439 | { |
d89f5eff JK |
440 | int r, i; |
441 | struct kvm *kvm = kvm_arch_alloc_vm(); | |
6aa8b732 | 442 | |
d89f5eff JK |
443 | if (!kvm) |
444 | return ERR_PTR(-ENOMEM); | |
445 | ||
446 | r = kvm_arch_init_vm(kvm); | |
447 | if (r) | |
448 | goto out_err_nodisable; | |
10474ae8 AG |
449 | |
450 | r = hardware_enable_all(); | |
451 | if (r) | |
452 | goto out_err_nodisable; | |
453 | ||
75858a84 AK |
454 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
455 | INIT_HLIST_HEAD(&kvm->mask_notifier_list); | |
136bdfee | 456 | INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list); |
75858a84 | 457 | #endif |
6aa8b732 | 458 | |
46a26bf5 MT |
459 | r = -ENOMEM; |
460 | kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
461 | if (!kvm->memslots) | |
57e7fbee | 462 | goto out_err_nosrcu; |
bc6678a3 | 463 | if (init_srcu_struct(&kvm->srcu)) |
57e7fbee | 464 | goto out_err_nosrcu; |
e93f8a0f MT |
465 | for (i = 0; i < KVM_NR_BUSES; i++) { |
466 | kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus), | |
467 | GFP_KERNEL); | |
57e7fbee | 468 | if (!kvm->buses[i]) |
e93f8a0f | 469 | goto out_err; |
e93f8a0f | 470 | } |
46a26bf5 | 471 | |
4c07b0a4 | 472 | r = kvm_init_mmu_notifier(kvm); |
57e7fbee | 473 | if (r) |
283d0c65 | 474 | goto out_err; |
e930bffe | 475 | |
6d4e4c4f AK |
476 | kvm->mm = current->mm; |
477 | atomic_inc(&kvm->mm->mm_count); | |
aaee2c94 | 478 | spin_lock_init(&kvm->mmu_lock); |
d34e6b17 | 479 | kvm_eventfd_init(kvm); |
11ec2804 | 480 | mutex_init(&kvm->lock); |
60eead79 | 481 | mutex_init(&kvm->irq_lock); |
79fac95e | 482 | mutex_init(&kvm->slots_lock); |
d39f13b0 | 483 | atomic_set(&kvm->users_count, 1); |
5e58cfe4 RR |
484 | spin_lock(&kvm_lock); |
485 | list_add(&kvm->vm_list, &vm_list); | |
486 | spin_unlock(&kvm_lock); | |
d89f5eff | 487 | |
f17abe9a | 488 | return kvm; |
10474ae8 AG |
489 | |
490 | out_err: | |
57e7fbee JK |
491 | cleanup_srcu_struct(&kvm->srcu); |
492 | out_err_nosrcu: | |
10474ae8 AG |
493 | hardware_disable_all(); |
494 | out_err_nodisable: | |
e93f8a0f MT |
495 | for (i = 0; i < KVM_NR_BUSES; i++) |
496 | kfree(kvm->buses[i]); | |
46a26bf5 | 497 | kfree(kvm->memslots); |
d89f5eff | 498 | kvm_arch_free_vm(kvm); |
10474ae8 | 499 | return ERR_PTR(r); |
f17abe9a AK |
500 | } |
501 | ||
a36a57b1 TY |
502 | static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot) |
503 | { | |
504 | if (!memslot->dirty_bitmap) | |
505 | return; | |
506 | ||
6f9e5c17 TY |
507 | if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE) |
508 | vfree(memslot->dirty_bitmap_head); | |
509 | else | |
510 | kfree(memslot->dirty_bitmap_head); | |
511 | ||
a36a57b1 | 512 | memslot->dirty_bitmap = NULL; |
515a0127 | 513 | memslot->dirty_bitmap_head = NULL; |
a36a57b1 TY |
514 | } |
515 | ||
6aa8b732 AK |
516 | /* |
517 | * Free any memory in @free but not in @dont. | |
518 | */ | |
519 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
520 | struct kvm_memory_slot *dont) | |
521 | { | |
ec04b260 JR |
522 | int i; |
523 | ||
290fc38d IE |
524 | if (!dont || free->rmap != dont->rmap) |
525 | vfree(free->rmap); | |
6aa8b732 AK |
526 | |
527 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
a36a57b1 | 528 | kvm_destroy_dirty_bitmap(free); |
6aa8b732 | 529 | |
ec04b260 JR |
530 | |
531 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { | |
532 | if (!dont || free->lpage_info[i] != dont->lpage_info[i]) { | |
533 | vfree(free->lpage_info[i]); | |
534 | free->lpage_info[i] = NULL; | |
535 | } | |
536 | } | |
05da4558 | 537 | |
6aa8b732 | 538 | free->npages = 0; |
8d4e1288 | 539 | free->rmap = NULL; |
6aa8b732 AK |
540 | } |
541 | ||
d19a9cd2 | 542 | void kvm_free_physmem(struct kvm *kvm) |
6aa8b732 AK |
543 | { |
544 | int i; | |
46a26bf5 MT |
545 | struct kvm_memslots *slots = kvm->memslots; |
546 | ||
547 | for (i = 0; i < slots->nmemslots; ++i) | |
548 | kvm_free_physmem_slot(&slots->memslots[i], NULL); | |
6aa8b732 | 549 | |
46a26bf5 | 550 | kfree(kvm->memslots); |
6aa8b732 AK |
551 | } |
552 | ||
f17abe9a AK |
553 | static void kvm_destroy_vm(struct kvm *kvm) |
554 | { | |
e93f8a0f | 555 | int i; |
6d4e4c4f AK |
556 | struct mm_struct *mm = kvm->mm; |
557 | ||
ad8ba2cd | 558 | kvm_arch_sync_events(kvm); |
133de902 AK |
559 | spin_lock(&kvm_lock); |
560 | list_del(&kvm->vm_list); | |
561 | spin_unlock(&kvm_lock); | |
399ec807 | 562 | kvm_free_irq_routing(kvm); |
e93f8a0f MT |
563 | for (i = 0; i < KVM_NR_BUSES; i++) |
564 | kvm_io_bus_destroy(kvm->buses[i]); | |
980da6ce | 565 | kvm_coalesced_mmio_free(kvm); |
e930bffe AA |
566 | #if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER) |
567 | mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm); | |
f00be0ca GN |
568 | #else |
569 | kvm_arch_flush_shadow(kvm); | |
5f94c174 | 570 | #endif |
d19a9cd2 | 571 | kvm_arch_destroy_vm(kvm); |
d89f5eff JK |
572 | kvm_free_physmem(kvm); |
573 | cleanup_srcu_struct(&kvm->srcu); | |
574 | kvm_arch_free_vm(kvm); | |
10474ae8 | 575 | hardware_disable_all(); |
6d4e4c4f | 576 | mmdrop(mm); |
f17abe9a AK |
577 | } |
578 | ||
d39f13b0 IE |
579 | void kvm_get_kvm(struct kvm *kvm) |
580 | { | |
581 | atomic_inc(&kvm->users_count); | |
582 | } | |
583 | EXPORT_SYMBOL_GPL(kvm_get_kvm); | |
584 | ||
585 | void kvm_put_kvm(struct kvm *kvm) | |
586 | { | |
587 | if (atomic_dec_and_test(&kvm->users_count)) | |
588 | kvm_destroy_vm(kvm); | |
589 | } | |
590 | EXPORT_SYMBOL_GPL(kvm_put_kvm); | |
591 | ||
592 | ||
f17abe9a AK |
593 | static int kvm_vm_release(struct inode *inode, struct file *filp) |
594 | { | |
595 | struct kvm *kvm = filp->private_data; | |
596 | ||
721eecbf GH |
597 | kvm_irqfd_release(kvm); |
598 | ||
d39f13b0 | 599 | kvm_put_kvm(kvm); |
6aa8b732 AK |
600 | return 0; |
601 | } | |
602 | ||
d48ead8b | 603 | #ifndef CONFIG_S390 |
515a0127 TY |
604 | /* |
605 | * Allocation size is twice as large as the actual dirty bitmap size. | |
606 | * This makes it possible to do double buffering: see x86's | |
607 | * kvm_vm_ioctl_get_dirty_log(). | |
608 | */ | |
a36a57b1 TY |
609 | static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot) |
610 | { | |
515a0127 | 611 | unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot); |
a36a57b1 | 612 | |
6f9e5c17 TY |
613 | if (dirty_bytes > PAGE_SIZE) |
614 | memslot->dirty_bitmap = vzalloc(dirty_bytes); | |
615 | else | |
616 | memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL); | |
617 | ||
a36a57b1 TY |
618 | if (!memslot->dirty_bitmap) |
619 | return -ENOMEM; | |
620 | ||
515a0127 | 621 | memslot->dirty_bitmap_head = memslot->dirty_bitmap; |
a36a57b1 TY |
622 | return 0; |
623 | } | |
d48ead8b | 624 | #endif /* !CONFIG_S390 */ |
a36a57b1 | 625 | |
6aa8b732 AK |
626 | /* |
627 | * Allocate some memory and give it an address in the guest physical address | |
628 | * space. | |
629 | * | |
630 | * Discontiguous memory is allowed, mostly for framebuffers. | |
f78e0e2e | 631 | * |
10589a46 | 632 | * Must be called holding mmap_sem for write. |
6aa8b732 | 633 | */ |
f78e0e2e SY |
634 | int __kvm_set_memory_region(struct kvm *kvm, |
635 | struct kvm_userspace_memory_region *mem, | |
636 | int user_alloc) | |
6aa8b732 | 637 | { |
8234b22e | 638 | int r; |
6aa8b732 | 639 | gfn_t base_gfn; |
28bcb112 HC |
640 | unsigned long npages; |
641 | unsigned long i; | |
6aa8b732 AK |
642 | struct kvm_memory_slot *memslot; |
643 | struct kvm_memory_slot old, new; | |
bc6678a3 | 644 | struct kvm_memslots *slots, *old_memslots; |
6aa8b732 AK |
645 | |
646 | r = -EINVAL; | |
647 | /* General sanity checks */ | |
648 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
649 | goto out; | |
650 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
651 | goto out; | |
e7cacd40 | 652 | if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1))) |
78749809 | 653 | goto out; |
e0d62c7f | 654 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
655 | goto out; |
656 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
657 | goto out; | |
658 | ||
46a26bf5 | 659 | memslot = &kvm->memslots->memslots[mem->slot]; |
6aa8b732 AK |
660 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; |
661 | npages = mem->memory_size >> PAGE_SHIFT; | |
662 | ||
660c22c4 TY |
663 | r = -EINVAL; |
664 | if (npages > KVM_MEM_MAX_NR_PAGES) | |
665 | goto out; | |
666 | ||
6aa8b732 AK |
667 | if (!npages) |
668 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
669 | ||
6aa8b732 AK |
670 | new = old = *memslot; |
671 | ||
e36d96f7 | 672 | new.id = mem->slot; |
6aa8b732 AK |
673 | new.base_gfn = base_gfn; |
674 | new.npages = npages; | |
675 | new.flags = mem->flags; | |
676 | ||
677 | /* Disallow changing a memory slot's size. */ | |
678 | r = -EINVAL; | |
679 | if (npages && old.npages && npages != old.npages) | |
f78e0e2e | 680 | goto out_free; |
6aa8b732 AK |
681 | |
682 | /* Check for overlaps */ | |
683 | r = -EEXIST; | |
684 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
46a26bf5 | 685 | struct kvm_memory_slot *s = &kvm->memslots->memslots[i]; |
6aa8b732 | 686 | |
4cd481f6 | 687 | if (s == memslot || !s->npages) |
6aa8b732 AK |
688 | continue; |
689 | if (!((base_gfn + npages <= s->base_gfn) || | |
690 | (base_gfn >= s->base_gfn + s->npages))) | |
f78e0e2e | 691 | goto out_free; |
6aa8b732 | 692 | } |
6aa8b732 | 693 | |
6aa8b732 AK |
694 | /* Free page dirty bitmap if unneeded */ |
695 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 696 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
697 | |
698 | r = -ENOMEM; | |
699 | ||
700 | /* Allocate if a slot is being created */ | |
eff0114a | 701 | #ifndef CONFIG_S390 |
8d4e1288 | 702 | if (npages && !new.rmap) { |
26535037 | 703 | new.rmap = vzalloc(npages * sizeof(*new.rmap)); |
290fc38d IE |
704 | |
705 | if (!new.rmap) | |
f78e0e2e | 706 | goto out_free; |
290fc38d | 707 | |
80b14b5b | 708 | new.user_alloc = user_alloc; |
bc6678a3 | 709 | new.userspace_addr = mem->userspace_addr; |
6aa8b732 | 710 | } |
ec04b260 JR |
711 | if (!npages) |
712 | goto skip_lpage; | |
05da4558 | 713 | |
ec04b260 | 714 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) { |
28bcb112 HC |
715 | unsigned long ugfn; |
716 | unsigned long j; | |
717 | int lpages; | |
ec04b260 | 718 | int level = i + 2; |
05da4558 | 719 | |
ec04b260 JR |
720 | /* Avoid unused variable warning if no large pages */ |
721 | (void)level; | |
722 | ||
723 | if (new.lpage_info[i]) | |
724 | continue; | |
725 | ||
82855413 JR |
726 | lpages = 1 + ((base_gfn + npages - 1) |
727 | >> KVM_HPAGE_GFN_SHIFT(level)); | |
728 | lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level); | |
ec04b260 | 729 | |
26535037 | 730 | new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i])); |
ec04b260 JR |
731 | |
732 | if (!new.lpage_info[i]) | |
05da4558 MT |
733 | goto out_free; |
734 | ||
82855413 | 735 | if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 736 | new.lpage_info[i][0].write_count = 1; |
82855413 | 737 | if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1)) |
ec04b260 | 738 | new.lpage_info[i][lpages - 1].write_count = 1; |
ac04527f AK |
739 | ugfn = new.userspace_addr >> PAGE_SHIFT; |
740 | /* | |
741 | * If the gfn and userspace address are not aligned wrt each | |
54dee993 MT |
742 | * other, or if explicitly asked to, disable large page |
743 | * support for this slot | |
ac04527f | 744 | */ |
ec04b260 | 745 | if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) || |
54dee993 | 746 | !largepages_enabled) |
ec04b260 JR |
747 | for (j = 0; j < lpages; ++j) |
748 | new.lpage_info[i][j].write_count = 1; | |
05da4558 | 749 | } |
6aa8b732 | 750 | |
ec04b260 JR |
751 | skip_lpage: |
752 | ||
6aa8b732 AK |
753 | /* Allocate page dirty bitmap if needed */ |
754 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
a36a57b1 | 755 | if (kvm_create_dirty_bitmap(&new) < 0) |
f78e0e2e | 756 | goto out_free; |
bc6678a3 | 757 | /* destroy any largepage mappings for dirty tracking */ |
6aa8b732 | 758 | } |
3eea8437 CB |
759 | #else /* not defined CONFIG_S390 */ |
760 | new.user_alloc = user_alloc; | |
761 | if (user_alloc) | |
762 | new.userspace_addr = mem->userspace_addr; | |
eff0114a | 763 | #endif /* not defined CONFIG_S390 */ |
6aa8b732 | 764 | |
bc6678a3 MT |
765 | if (!npages) { |
766 | r = -ENOMEM; | |
767 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
768 | if (!slots) | |
769 | goto out_free; | |
770 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
771 | if (mem->slot >= slots->nmemslots) | |
772 | slots->nmemslots = mem->slot + 1; | |
49c7754c | 773 | slots->generation++; |
bc6678a3 MT |
774 | slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID; |
775 | ||
776 | old_memslots = kvm->memslots; | |
777 | rcu_assign_pointer(kvm->memslots, slots); | |
778 | synchronize_srcu_expedited(&kvm->srcu); | |
779 | /* From this point no new shadow pages pointing to a deleted | |
780 | * memslot will be created. | |
781 | * | |
782 | * validation of sp->gfn happens in: | |
783 | * - gfn_to_hva (kvm_read_guest, gfn_to_pfn) | |
784 | * - kvm_is_visible_gfn (mmu_check_roots) | |
785 | */ | |
34d4cb8f | 786 | kvm_arch_flush_shadow(kvm); |
bc6678a3 MT |
787 | kfree(old_memslots); |
788 | } | |
34d4cb8f | 789 | |
f7784b8e MT |
790 | r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc); |
791 | if (r) | |
792 | goto out_free; | |
793 | ||
bc6678a3 MT |
794 | /* map the pages in iommu page table */ |
795 | if (npages) { | |
796 | r = kvm_iommu_map_pages(kvm, &new); | |
797 | if (r) | |
798 | goto out_free; | |
799 | } | |
604b38ac | 800 | |
bc6678a3 MT |
801 | r = -ENOMEM; |
802 | slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL); | |
803 | if (!slots) | |
804 | goto out_free; | |
805 | memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots)); | |
806 | if (mem->slot >= slots->nmemslots) | |
807 | slots->nmemslots = mem->slot + 1; | |
49c7754c | 808 | slots->generation++; |
bc6678a3 MT |
809 | |
810 | /* actual memory is freed via old in kvm_free_physmem_slot below */ | |
811 | if (!npages) { | |
812 | new.rmap = NULL; | |
813 | new.dirty_bitmap = NULL; | |
814 | for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) | |
815 | new.lpage_info[i] = NULL; | |
816 | } | |
817 | ||
818 | slots->memslots[mem->slot] = new; | |
819 | old_memslots = kvm->memslots; | |
820 | rcu_assign_pointer(kvm->memslots, slots); | |
821 | synchronize_srcu_expedited(&kvm->srcu); | |
3ad82a7e | 822 | |
f7784b8e | 823 | kvm_arch_commit_memory_region(kvm, mem, old, user_alloc); |
82ce2c96 | 824 | |
bc6678a3 MT |
825 | kvm_free_physmem_slot(&old, &new); |
826 | kfree(old_memslots); | |
827 | ||
6aa8b732 AK |
828 | return 0; |
829 | ||
f78e0e2e | 830 | out_free: |
6aa8b732 AK |
831 | kvm_free_physmem_slot(&new, &old); |
832 | out: | |
833 | return r; | |
210c7c4d IE |
834 | |
835 | } | |
f78e0e2e SY |
836 | EXPORT_SYMBOL_GPL(__kvm_set_memory_region); |
837 | ||
838 | int kvm_set_memory_region(struct kvm *kvm, | |
839 | struct kvm_userspace_memory_region *mem, | |
840 | int user_alloc) | |
841 | { | |
842 | int r; | |
843 | ||
79fac95e | 844 | mutex_lock(&kvm->slots_lock); |
f78e0e2e | 845 | r = __kvm_set_memory_region(kvm, mem, user_alloc); |
79fac95e | 846 | mutex_unlock(&kvm->slots_lock); |
f78e0e2e SY |
847 | return r; |
848 | } | |
210c7c4d IE |
849 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); |
850 | ||
1fe779f8 CO |
851 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
852 | struct | |
853 | kvm_userspace_memory_region *mem, | |
854 | int user_alloc) | |
210c7c4d | 855 | { |
e0d62c7f IE |
856 | if (mem->slot >= KVM_MEMORY_SLOTS) |
857 | return -EINVAL; | |
210c7c4d | 858 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
859 | } |
860 | ||
5bb064dc ZX |
861 | int kvm_get_dirty_log(struct kvm *kvm, |
862 | struct kvm_dirty_log *log, int *is_dirty) | |
6aa8b732 AK |
863 | { |
864 | struct kvm_memory_slot *memslot; | |
865 | int r, i; | |
87bf6e7d | 866 | unsigned long n; |
6aa8b732 AK |
867 | unsigned long any = 0; |
868 | ||
6aa8b732 AK |
869 | r = -EINVAL; |
870 | if (log->slot >= KVM_MEMORY_SLOTS) | |
871 | goto out; | |
872 | ||
46a26bf5 | 873 | memslot = &kvm->memslots->memslots[log->slot]; |
6aa8b732 AK |
874 | r = -ENOENT; |
875 | if (!memslot->dirty_bitmap) | |
876 | goto out; | |
877 | ||
87bf6e7d | 878 | n = kvm_dirty_bitmap_bytes(memslot); |
6aa8b732 | 879 | |
cd1a4a98 | 880 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
881 | any = memslot->dirty_bitmap[i]; |
882 | ||
883 | r = -EFAULT; | |
884 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
885 | goto out; | |
886 | ||
5bb064dc ZX |
887 | if (any) |
888 | *is_dirty = 1; | |
6aa8b732 AK |
889 | |
890 | r = 0; | |
6aa8b732 | 891 | out: |
6aa8b732 AK |
892 | return r; |
893 | } | |
894 | ||
54dee993 MT |
895 | void kvm_disable_largepages(void) |
896 | { | |
897 | largepages_enabled = false; | |
898 | } | |
899 | EXPORT_SYMBOL_GPL(kvm_disable_largepages); | |
900 | ||
cea7bb21 IE |
901 | int is_error_page(struct page *page) |
902 | { | |
edba23e5 | 903 | return page == bad_page || page == hwpoison_page || page == fault_page; |
cea7bb21 IE |
904 | } |
905 | EXPORT_SYMBOL_GPL(is_error_page); | |
906 | ||
35149e21 AL |
907 | int is_error_pfn(pfn_t pfn) |
908 | { | |
edba23e5 | 909 | return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn; |
35149e21 AL |
910 | } |
911 | EXPORT_SYMBOL_GPL(is_error_pfn); | |
912 | ||
bf998156 HY |
913 | int is_hwpoison_pfn(pfn_t pfn) |
914 | { | |
915 | return pfn == hwpoison_pfn; | |
916 | } | |
917 | EXPORT_SYMBOL_GPL(is_hwpoison_pfn); | |
918 | ||
edba23e5 GN |
919 | int is_fault_pfn(pfn_t pfn) |
920 | { | |
921 | return pfn == fault_pfn; | |
922 | } | |
923 | EXPORT_SYMBOL_GPL(is_fault_pfn); | |
924 | ||
f9d46eb0 IE |
925 | static inline unsigned long bad_hva(void) |
926 | { | |
927 | return PAGE_OFFSET; | |
928 | } | |
929 | ||
930 | int kvm_is_error_hva(unsigned long addr) | |
931 | { | |
932 | return addr == bad_hva(); | |
933 | } | |
934 | EXPORT_SYMBOL_GPL(kvm_is_error_hva); | |
935 | ||
49c7754c GN |
936 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots, |
937 | gfn_t gfn) | |
6aa8b732 AK |
938 | { |
939 | int i; | |
940 | ||
46a26bf5 MT |
941 | for (i = 0; i < slots->nmemslots; ++i) { |
942 | struct kvm_memory_slot *memslot = &slots->memslots[i]; | |
6aa8b732 AK |
943 | |
944 | if (gfn >= memslot->base_gfn | |
945 | && gfn < memslot->base_gfn + memslot->npages) | |
946 | return memslot; | |
947 | } | |
8b6d44c7 | 948 | return NULL; |
6aa8b732 | 949 | } |
49c7754c GN |
950 | |
951 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
952 | { | |
953 | return __gfn_to_memslot(kvm_memslots(kvm), gfn); | |
954 | } | |
a1f4d395 | 955 | EXPORT_SYMBOL_GPL(gfn_to_memslot); |
6aa8b732 | 956 | |
e0d62c7f IE |
957 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
958 | { | |
959 | int i; | |
90d83dc3 | 960 | struct kvm_memslots *slots = kvm_memslots(kvm); |
e0d62c7f | 961 | |
e0d62c7f | 962 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { |
46a26bf5 | 963 | struct kvm_memory_slot *memslot = &slots->memslots[i]; |
e0d62c7f | 964 | |
bc6678a3 MT |
965 | if (memslot->flags & KVM_MEMSLOT_INVALID) |
966 | continue; | |
967 | ||
e0d62c7f IE |
968 | if (gfn >= memslot->base_gfn |
969 | && gfn < memslot->base_gfn + memslot->npages) | |
970 | return 1; | |
971 | } | |
972 | return 0; | |
973 | } | |
974 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
975 | ||
8f0b1ab6 JR |
976 | unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn) |
977 | { | |
978 | struct vm_area_struct *vma; | |
979 | unsigned long addr, size; | |
980 | ||
981 | size = PAGE_SIZE; | |
982 | ||
983 | addr = gfn_to_hva(kvm, gfn); | |
984 | if (kvm_is_error_hva(addr)) | |
985 | return PAGE_SIZE; | |
986 | ||
987 | down_read(¤t->mm->mmap_sem); | |
988 | vma = find_vma(current->mm, addr); | |
989 | if (!vma) | |
990 | goto out; | |
991 | ||
992 | size = vma_kernel_pagesize(vma); | |
993 | ||
994 | out: | |
995 | up_read(¤t->mm->mmap_sem); | |
996 | ||
997 | return size; | |
998 | } | |
999 | ||
bc6678a3 MT |
1000 | int memslot_id(struct kvm *kvm, gfn_t gfn) |
1001 | { | |
1002 | int i; | |
90d83dc3 | 1003 | struct kvm_memslots *slots = kvm_memslots(kvm); |
bc6678a3 MT |
1004 | struct kvm_memory_slot *memslot = NULL; |
1005 | ||
bc6678a3 MT |
1006 | for (i = 0; i < slots->nmemslots; ++i) { |
1007 | memslot = &slots->memslots[i]; | |
1008 | ||
1009 | if (gfn >= memslot->base_gfn | |
1010 | && gfn < memslot->base_gfn + memslot->npages) | |
1011 | break; | |
1012 | } | |
1013 | ||
1014 | return memslot - slots->memslots; | |
1015 | } | |
1016 | ||
49c7754c | 1017 | static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn, |
48987781 | 1018 | gfn_t *nr_pages) |
539cb660 | 1019 | { |
bc6678a3 | 1020 | if (!slot || slot->flags & KVM_MEMSLOT_INVALID) |
539cb660 | 1021 | return bad_hva(); |
48987781 XG |
1022 | |
1023 | if (nr_pages) | |
1024 | *nr_pages = slot->npages - (gfn - slot->base_gfn); | |
1025 | ||
f5c98031 | 1026 | return gfn_to_hva_memslot(slot, gfn); |
539cb660 | 1027 | } |
48987781 XG |
1028 | |
1029 | unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn) | |
1030 | { | |
49c7754c | 1031 | return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL); |
48987781 | 1032 | } |
0d150298 | 1033 | EXPORT_SYMBOL_GPL(gfn_to_hva); |
539cb660 | 1034 | |
8030089f GN |
1035 | static pfn_t get_fault_pfn(void) |
1036 | { | |
1037 | get_page(fault_page); | |
1038 | return fault_pfn; | |
1039 | } | |
1040 | ||
fafc3dba HY |
1041 | static inline int check_user_page_hwpoison(unsigned long addr) |
1042 | { | |
1043 | int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE; | |
1044 | ||
1045 | rc = __get_user_pages(current, current->mm, addr, 1, | |
1046 | flags, NULL, NULL, NULL); | |
1047 | return rc == -EHWPOISON; | |
1048 | } | |
1049 | ||
af585b92 | 1050 | static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic, |
612819c3 | 1051 | bool *async, bool write_fault, bool *writable) |
954bbbc2 | 1052 | { |
8d4e1288 | 1053 | struct page *page[1]; |
af585b92 | 1054 | int npages = 0; |
2e2e3738 | 1055 | pfn_t pfn; |
954bbbc2 | 1056 | |
af585b92 GN |
1057 | /* we can do it either atomically or asynchronously, not both */ |
1058 | BUG_ON(atomic && async); | |
1059 | ||
612819c3 MT |
1060 | BUG_ON(!write_fault && !writable); |
1061 | ||
1062 | if (writable) | |
1063 | *writable = true; | |
1064 | ||
af585b92 | 1065 | if (atomic || async) |
887c08ac | 1066 | npages = __get_user_pages_fast(addr, 1, 1, page); |
af585b92 GN |
1067 | |
1068 | if (unlikely(npages != 1) && !atomic) { | |
887c08ac | 1069 | might_sleep(); |
612819c3 MT |
1070 | |
1071 | if (writable) | |
1072 | *writable = write_fault; | |
1073 | ||
1074 | npages = get_user_pages_fast(addr, 1, write_fault, page); | |
1075 | ||
1076 | /* map read fault as writable if possible */ | |
1077 | if (unlikely(!write_fault) && npages == 1) { | |
1078 | struct page *wpage[1]; | |
1079 | ||
1080 | npages = __get_user_pages_fast(addr, 1, 1, wpage); | |
1081 | if (npages == 1) { | |
1082 | *writable = true; | |
1083 | put_page(page[0]); | |
1084 | page[0] = wpage[0]; | |
1085 | } | |
1086 | npages = 1; | |
1087 | } | |
887c08ac | 1088 | } |
539cb660 | 1089 | |
2e2e3738 AL |
1090 | if (unlikely(npages != 1)) { |
1091 | struct vm_area_struct *vma; | |
1092 | ||
887c08ac | 1093 | if (atomic) |
8030089f | 1094 | return get_fault_pfn(); |
887c08ac | 1095 | |
bbeb3406 | 1096 | down_read(¤t->mm->mmap_sem); |
fafc3dba | 1097 | if (check_user_page_hwpoison(addr)) { |
bbeb3406 | 1098 | up_read(¤t->mm->mmap_sem); |
bf998156 HY |
1099 | get_page(hwpoison_page); |
1100 | return page_to_pfn(hwpoison_page); | |
1101 | } | |
1102 | ||
8030089f | 1103 | vma = find_vma_intersection(current->mm, addr, addr+1); |
4c2155ce | 1104 | |
8030089f GN |
1105 | if (vma == NULL) |
1106 | pfn = get_fault_pfn(); | |
1107 | else if ((vma->vm_flags & VM_PFNMAP)) { | |
1108 | pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + | |
1109 | vma->vm_pgoff; | |
1110 | BUG_ON(!kvm_is_mmio_pfn(pfn)); | |
1111 | } else { | |
1112 | if (async && (vma->vm_flags & VM_WRITE)) | |
af585b92 | 1113 | *async = true; |
8030089f | 1114 | pfn = get_fault_pfn(); |
2e2e3738 | 1115 | } |
4c2155ce | 1116 | up_read(¤t->mm->mmap_sem); |
2e2e3738 AL |
1117 | } else |
1118 | pfn = page_to_pfn(page[0]); | |
8d4e1288 | 1119 | |
2e2e3738 | 1120 | return pfn; |
35149e21 AL |
1121 | } |
1122 | ||
887c08ac XG |
1123 | pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr) |
1124 | { | |
612819c3 | 1125 | return hva_to_pfn(kvm, addr, true, NULL, true, NULL); |
887c08ac XG |
1126 | } |
1127 | EXPORT_SYMBOL_GPL(hva_to_pfn_atomic); | |
1128 | ||
612819c3 MT |
1129 | static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async, |
1130 | bool write_fault, bool *writable) | |
506f0d6f MT |
1131 | { |
1132 | unsigned long addr; | |
1133 | ||
af585b92 GN |
1134 | if (async) |
1135 | *async = false; | |
1136 | ||
506f0d6f MT |
1137 | addr = gfn_to_hva(kvm, gfn); |
1138 | if (kvm_is_error_hva(addr)) { | |
1139 | get_page(bad_page); | |
1140 | return page_to_pfn(bad_page); | |
1141 | } | |
1142 | ||
612819c3 | 1143 | return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable); |
365fb3fd XG |
1144 | } |
1145 | ||
1146 | pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn) | |
1147 | { | |
612819c3 | 1148 | return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL); |
365fb3fd XG |
1149 | } |
1150 | EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic); | |
1151 | ||
612819c3 MT |
1152 | pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async, |
1153 | bool write_fault, bool *writable) | |
af585b92 | 1154 | { |
612819c3 | 1155 | return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable); |
af585b92 GN |
1156 | } |
1157 | EXPORT_SYMBOL_GPL(gfn_to_pfn_async); | |
1158 | ||
365fb3fd XG |
1159 | pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn) |
1160 | { | |
612819c3 | 1161 | return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL); |
506f0d6f | 1162 | } |
35149e21 AL |
1163 | EXPORT_SYMBOL_GPL(gfn_to_pfn); |
1164 | ||
612819c3 MT |
1165 | pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault, |
1166 | bool *writable) | |
1167 | { | |
1168 | return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable); | |
1169 | } | |
1170 | EXPORT_SYMBOL_GPL(gfn_to_pfn_prot); | |
1171 | ||
506f0d6f MT |
1172 | pfn_t gfn_to_pfn_memslot(struct kvm *kvm, |
1173 | struct kvm_memory_slot *slot, gfn_t gfn) | |
1174 | { | |
1175 | unsigned long addr = gfn_to_hva_memslot(slot, gfn); | |
612819c3 | 1176 | return hva_to_pfn(kvm, addr, false, NULL, true, NULL); |
506f0d6f MT |
1177 | } |
1178 | ||
48987781 XG |
1179 | int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages, |
1180 | int nr_pages) | |
1181 | { | |
1182 | unsigned long addr; | |
1183 | gfn_t entry; | |
1184 | ||
49c7754c | 1185 | addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry); |
48987781 XG |
1186 | if (kvm_is_error_hva(addr)) |
1187 | return -1; | |
1188 | ||
1189 | if (entry < nr_pages) | |
1190 | return 0; | |
1191 | ||
1192 | return __get_user_pages_fast(addr, nr_pages, 1, pages); | |
1193 | } | |
1194 | EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic); | |
1195 | ||
35149e21 AL |
1196 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
1197 | { | |
2e2e3738 AL |
1198 | pfn_t pfn; |
1199 | ||
1200 | pfn = gfn_to_pfn(kvm, gfn); | |
c77fb9dc | 1201 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 AL |
1202 | return pfn_to_page(pfn); |
1203 | ||
c77fb9dc | 1204 | WARN_ON(kvm_is_mmio_pfn(pfn)); |
2e2e3738 AL |
1205 | |
1206 | get_page(bad_page); | |
1207 | return bad_page; | |
954bbbc2 | 1208 | } |
aab61cc0 | 1209 | |
954bbbc2 AK |
1210 | EXPORT_SYMBOL_GPL(gfn_to_page); |
1211 | ||
b4231d61 IE |
1212 | void kvm_release_page_clean(struct page *page) |
1213 | { | |
35149e21 | 1214 | kvm_release_pfn_clean(page_to_pfn(page)); |
b4231d61 IE |
1215 | } |
1216 | EXPORT_SYMBOL_GPL(kvm_release_page_clean); | |
1217 | ||
35149e21 AL |
1218 | void kvm_release_pfn_clean(pfn_t pfn) |
1219 | { | |
c77fb9dc | 1220 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1221 | put_page(pfn_to_page(pfn)); |
35149e21 AL |
1222 | } |
1223 | EXPORT_SYMBOL_GPL(kvm_release_pfn_clean); | |
1224 | ||
b4231d61 | 1225 | void kvm_release_page_dirty(struct page *page) |
8a7ae055 | 1226 | { |
35149e21 AL |
1227 | kvm_release_pfn_dirty(page_to_pfn(page)); |
1228 | } | |
1229 | EXPORT_SYMBOL_GPL(kvm_release_page_dirty); | |
1230 | ||
1231 | void kvm_release_pfn_dirty(pfn_t pfn) | |
1232 | { | |
1233 | kvm_set_pfn_dirty(pfn); | |
1234 | kvm_release_pfn_clean(pfn); | |
1235 | } | |
1236 | EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty); | |
1237 | ||
1238 | void kvm_set_page_dirty(struct page *page) | |
1239 | { | |
1240 | kvm_set_pfn_dirty(page_to_pfn(page)); | |
1241 | } | |
1242 | EXPORT_SYMBOL_GPL(kvm_set_page_dirty); | |
1243 | ||
1244 | void kvm_set_pfn_dirty(pfn_t pfn) | |
1245 | { | |
c77fb9dc | 1246 | if (!kvm_is_mmio_pfn(pfn)) { |
2e2e3738 AL |
1247 | struct page *page = pfn_to_page(pfn); |
1248 | if (!PageReserved(page)) | |
1249 | SetPageDirty(page); | |
1250 | } | |
8a7ae055 | 1251 | } |
35149e21 AL |
1252 | EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty); |
1253 | ||
1254 | void kvm_set_pfn_accessed(pfn_t pfn) | |
1255 | { | |
c77fb9dc | 1256 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1257 | mark_page_accessed(pfn_to_page(pfn)); |
35149e21 AL |
1258 | } |
1259 | EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed); | |
1260 | ||
1261 | void kvm_get_pfn(pfn_t pfn) | |
1262 | { | |
c77fb9dc | 1263 | if (!kvm_is_mmio_pfn(pfn)) |
2e2e3738 | 1264 | get_page(pfn_to_page(pfn)); |
35149e21 AL |
1265 | } |
1266 | EXPORT_SYMBOL_GPL(kvm_get_pfn); | |
8a7ae055 | 1267 | |
195aefde IE |
1268 | static int next_segment(unsigned long len, int offset) |
1269 | { | |
1270 | if (len > PAGE_SIZE - offset) | |
1271 | return PAGE_SIZE - offset; | |
1272 | else | |
1273 | return len; | |
1274 | } | |
1275 | ||
1276 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
1277 | int len) | |
1278 | { | |
e0506bcb IE |
1279 | int r; |
1280 | unsigned long addr; | |
195aefde | 1281 | |
e0506bcb IE |
1282 | addr = gfn_to_hva(kvm, gfn); |
1283 | if (kvm_is_error_hva(addr)) | |
1284 | return -EFAULT; | |
1285 | r = copy_from_user(data, (void __user *)addr + offset, len); | |
1286 | if (r) | |
195aefde | 1287 | return -EFAULT; |
195aefde IE |
1288 | return 0; |
1289 | } | |
1290 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
1291 | ||
1292 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
1293 | { | |
1294 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1295 | int seg; | |
1296 | int offset = offset_in_page(gpa); | |
1297 | int ret; | |
1298 | ||
1299 | while ((seg = next_segment(len, offset)) != 0) { | |
1300 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
1301 | if (ret < 0) | |
1302 | return ret; | |
1303 | offset = 0; | |
1304 | len -= seg; | |
1305 | data += seg; | |
1306 | ++gfn; | |
1307 | } | |
1308 | return 0; | |
1309 | } | |
1310 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
1311 | ||
7ec54588 MT |
1312 | int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data, |
1313 | unsigned long len) | |
1314 | { | |
1315 | int r; | |
1316 | unsigned long addr; | |
1317 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1318 | int offset = offset_in_page(gpa); | |
1319 | ||
1320 | addr = gfn_to_hva(kvm, gfn); | |
1321 | if (kvm_is_error_hva(addr)) | |
1322 | return -EFAULT; | |
0aac03f0 | 1323 | pagefault_disable(); |
7ec54588 | 1324 | r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len); |
0aac03f0 | 1325 | pagefault_enable(); |
7ec54588 MT |
1326 | if (r) |
1327 | return -EFAULT; | |
1328 | return 0; | |
1329 | } | |
1330 | EXPORT_SYMBOL(kvm_read_guest_atomic); | |
1331 | ||
195aefde IE |
1332 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, |
1333 | int offset, int len) | |
1334 | { | |
e0506bcb IE |
1335 | int r; |
1336 | unsigned long addr; | |
195aefde | 1337 | |
e0506bcb IE |
1338 | addr = gfn_to_hva(kvm, gfn); |
1339 | if (kvm_is_error_hva(addr)) | |
1340 | return -EFAULT; | |
1341 | r = copy_to_user((void __user *)addr + offset, data, len); | |
1342 | if (r) | |
195aefde | 1343 | return -EFAULT; |
195aefde IE |
1344 | mark_page_dirty(kvm, gfn); |
1345 | return 0; | |
1346 | } | |
1347 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
1348 | ||
1349 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
1350 | unsigned long len) | |
1351 | { | |
1352 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1353 | int seg; | |
1354 | int offset = offset_in_page(gpa); | |
1355 | int ret; | |
1356 | ||
1357 | while ((seg = next_segment(len, offset)) != 0) { | |
1358 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
1359 | if (ret < 0) | |
1360 | return ret; | |
1361 | offset = 0; | |
1362 | len -= seg; | |
1363 | data += seg; | |
1364 | ++gfn; | |
1365 | } | |
1366 | return 0; | |
1367 | } | |
1368 | ||
49c7754c GN |
1369 | int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc, |
1370 | gpa_t gpa) | |
1371 | { | |
1372 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1373 | int offset = offset_in_page(gpa); | |
1374 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1375 | ||
1376 | ghc->gpa = gpa; | |
1377 | ghc->generation = slots->generation; | |
1378 | ghc->memslot = __gfn_to_memslot(slots, gfn); | |
1379 | ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL); | |
1380 | if (!kvm_is_error_hva(ghc->hva)) | |
1381 | ghc->hva += offset; | |
1382 | else | |
1383 | return -EFAULT; | |
1384 | ||
1385 | return 0; | |
1386 | } | |
1387 | EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init); | |
1388 | ||
1389 | int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc, | |
1390 | void *data, unsigned long len) | |
1391 | { | |
1392 | struct kvm_memslots *slots = kvm_memslots(kvm); | |
1393 | int r; | |
1394 | ||
1395 | if (slots->generation != ghc->generation) | |
1396 | kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa); | |
1397 | ||
1398 | if (kvm_is_error_hva(ghc->hva)) | |
1399 | return -EFAULT; | |
1400 | ||
1401 | r = copy_to_user((void __user *)ghc->hva, data, len); | |
1402 | if (r) | |
1403 | return -EFAULT; | |
1404 | mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT); | |
1405 | ||
1406 | return 0; | |
1407 | } | |
1408 | EXPORT_SYMBOL_GPL(kvm_write_guest_cached); | |
1409 | ||
195aefde IE |
1410 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) |
1411 | { | |
3bcc8a8c HC |
1412 | return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page, |
1413 | offset, len); | |
195aefde IE |
1414 | } |
1415 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
1416 | ||
1417 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
1418 | { | |
1419 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
1420 | int seg; | |
1421 | int offset = offset_in_page(gpa); | |
1422 | int ret; | |
1423 | ||
1424 | while ((seg = next_segment(len, offset)) != 0) { | |
1425 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
1426 | if (ret < 0) | |
1427 | return ret; | |
1428 | offset = 0; | |
1429 | len -= seg; | |
1430 | ++gfn; | |
1431 | } | |
1432 | return 0; | |
1433 | } | |
1434 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
1435 | ||
49c7754c GN |
1436 | void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot, |
1437 | gfn_t gfn) | |
6aa8b732 | 1438 | { |
7e9d619d RR |
1439 | if (memslot && memslot->dirty_bitmap) { |
1440 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1441 | |
d1476937 | 1442 | generic___set_le_bit(rel_gfn, memslot->dirty_bitmap); |
6aa8b732 AK |
1443 | } |
1444 | } | |
1445 | ||
49c7754c GN |
1446 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1447 | { | |
1448 | struct kvm_memory_slot *memslot; | |
1449 | ||
1450 | memslot = gfn_to_memslot(kvm, gfn); | |
1451 | mark_page_dirty_in_slot(kvm, memslot, gfn); | |
1452 | } | |
1453 | ||
b6958ce4 ED |
1454 | /* |
1455 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1456 | */ | |
8776e519 | 1457 | void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1458 | { |
e5c239cf MT |
1459 | DEFINE_WAIT(wait); |
1460 | ||
1461 | for (;;) { | |
1462 | prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE); | |
1463 | ||
a1b37100 | 1464 | if (kvm_arch_vcpu_runnable(vcpu)) { |
a8eeb04a | 1465 | kvm_make_request(KVM_REQ_UNHALT, vcpu); |
e5c239cf | 1466 | break; |
d7690175 | 1467 | } |
09cec754 GN |
1468 | if (kvm_cpu_has_pending_timer(vcpu)) |
1469 | break; | |
e5c239cf MT |
1470 | if (signal_pending(current)) |
1471 | break; | |
1472 | ||
b6958ce4 | 1473 | schedule(); |
b6958ce4 | 1474 | } |
d3bef15f | 1475 | |
e5c239cf | 1476 | finish_wait(&vcpu->wq, &wait); |
b6958ce4 ED |
1477 | } |
1478 | ||
6aa8b732 AK |
1479 | void kvm_resched(struct kvm_vcpu *vcpu) |
1480 | { | |
3fca0365 YD |
1481 | if (!need_resched()) |
1482 | return; | |
6aa8b732 | 1483 | cond_resched(); |
6aa8b732 AK |
1484 | } |
1485 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1486 | ||
217ece61 | 1487 | void kvm_vcpu_on_spin(struct kvm_vcpu *me) |
d255f4f2 | 1488 | { |
217ece61 RR |
1489 | struct kvm *kvm = me->kvm; |
1490 | struct kvm_vcpu *vcpu; | |
1491 | int last_boosted_vcpu = me->kvm->last_boosted_vcpu; | |
1492 | int yielded = 0; | |
1493 | int pass; | |
1494 | int i; | |
d255f4f2 | 1495 | |
217ece61 RR |
1496 | /* |
1497 | * We boost the priority of a VCPU that is runnable but not | |
1498 | * currently running, because it got preempted by something | |
1499 | * else and called schedule in __vcpu_run. Hopefully that | |
1500 | * VCPU is holding the lock that we need and will release it. | |
1501 | * We approximate round-robin by starting at the last boosted VCPU. | |
1502 | */ | |
1503 | for (pass = 0; pass < 2 && !yielded; pass++) { | |
1504 | kvm_for_each_vcpu(i, vcpu, kvm) { | |
1505 | struct task_struct *task = NULL; | |
1506 | struct pid *pid; | |
1507 | if (!pass && i < last_boosted_vcpu) { | |
1508 | i = last_boosted_vcpu; | |
1509 | continue; | |
1510 | } else if (pass && i > last_boosted_vcpu) | |
1511 | break; | |
1512 | if (vcpu == me) | |
1513 | continue; | |
1514 | if (waitqueue_active(&vcpu->wq)) | |
1515 | continue; | |
1516 | rcu_read_lock(); | |
1517 | pid = rcu_dereference(vcpu->pid); | |
1518 | if (pid) | |
1519 | task = get_pid_task(vcpu->pid, PIDTYPE_PID); | |
1520 | rcu_read_unlock(); | |
1521 | if (!task) | |
1522 | continue; | |
1523 | if (task->flags & PF_VCPU) { | |
1524 | put_task_struct(task); | |
1525 | continue; | |
1526 | } | |
1527 | if (yield_to(task, 1)) { | |
1528 | put_task_struct(task); | |
1529 | kvm->last_boosted_vcpu = i; | |
1530 | yielded = 1; | |
1531 | break; | |
1532 | } | |
1533 | put_task_struct(task); | |
1534 | } | |
1535 | } | |
d255f4f2 ZE |
1536 | } |
1537 | EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin); | |
1538 | ||
e4a533a4 | 1539 | static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
9a2bb7f4 AK |
1540 | { |
1541 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
9a2bb7f4 AK |
1542 | struct page *page; |
1543 | ||
e4a533a4 | 1544 | if (vmf->pgoff == 0) |
039576c0 | 1545 | page = virt_to_page(vcpu->run); |
09566765 | 1546 | #ifdef CONFIG_X86 |
e4a533a4 | 1547 | else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET) |
ad312c7c | 1548 | page = virt_to_page(vcpu->arch.pio_data); |
5f94c174 LV |
1549 | #endif |
1550 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
1551 | else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET) | |
1552 | page = virt_to_page(vcpu->kvm->coalesced_mmio_ring); | |
09566765 | 1553 | #endif |
039576c0 | 1554 | else |
e4a533a4 | 1555 | return VM_FAULT_SIGBUS; |
9a2bb7f4 | 1556 | get_page(page); |
e4a533a4 | 1557 | vmf->page = page; |
1558 | return 0; | |
9a2bb7f4 AK |
1559 | } |
1560 | ||
f0f37e2f | 1561 | static const struct vm_operations_struct kvm_vcpu_vm_ops = { |
e4a533a4 | 1562 | .fault = kvm_vcpu_fault, |
9a2bb7f4 AK |
1563 | }; |
1564 | ||
1565 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
1566 | { | |
1567 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
1568 | return 0; | |
1569 | } | |
1570 | ||
bccf2150 AK |
1571 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
1572 | { | |
1573 | struct kvm_vcpu *vcpu = filp->private_data; | |
1574 | ||
66c0b394 | 1575 | kvm_put_kvm(vcpu->kvm); |
bccf2150 AK |
1576 | return 0; |
1577 | } | |
1578 | ||
3d3aab1b | 1579 | static struct file_operations kvm_vcpu_fops = { |
bccf2150 AK |
1580 | .release = kvm_vcpu_release, |
1581 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
1582 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 1583 | .mmap = kvm_vcpu_mmap, |
6038f373 | 1584 | .llseek = noop_llseek, |
bccf2150 AK |
1585 | }; |
1586 | ||
1587 | /* | |
1588 | * Allocates an inode for the vcpu. | |
1589 | */ | |
1590 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
1591 | { | |
628ff7c1 | 1592 | return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR); |
bccf2150 AK |
1593 | } |
1594 | ||
c5ea7660 AK |
1595 | /* |
1596 | * Creates some virtual cpus. Good luck creating more than one. | |
1597 | */ | |
73880c80 | 1598 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id) |
c5ea7660 AK |
1599 | { |
1600 | int r; | |
988a2cae | 1601 | struct kvm_vcpu *vcpu, *v; |
c5ea7660 | 1602 | |
73880c80 | 1603 | vcpu = kvm_arch_vcpu_create(kvm, id); |
fb3f0f51 RR |
1604 | if (IS_ERR(vcpu)) |
1605 | return PTR_ERR(vcpu); | |
c5ea7660 | 1606 | |
15ad7146 AK |
1607 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
1608 | ||
26e5215f AK |
1609 | r = kvm_arch_vcpu_setup(vcpu); |
1610 | if (r) | |
7d8fece6 | 1611 | return r; |
26e5215f | 1612 | |
11ec2804 | 1613 | mutex_lock(&kvm->lock); |
73880c80 GN |
1614 | if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) { |
1615 | r = -EINVAL; | |
e9b11c17 | 1616 | goto vcpu_destroy; |
fb3f0f51 | 1617 | } |
73880c80 | 1618 | |
988a2cae GN |
1619 | kvm_for_each_vcpu(r, v, kvm) |
1620 | if (v->vcpu_id == id) { | |
73880c80 GN |
1621 | r = -EEXIST; |
1622 | goto vcpu_destroy; | |
1623 | } | |
1624 | ||
1625 | BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]); | |
c5ea7660 | 1626 | |
fb3f0f51 | 1627 | /* Now it's all set up, let userspace reach it */ |
66c0b394 | 1628 | kvm_get_kvm(kvm); |
bccf2150 | 1629 | r = create_vcpu_fd(vcpu); |
73880c80 GN |
1630 | if (r < 0) { |
1631 | kvm_put_kvm(kvm); | |
1632 | goto vcpu_destroy; | |
1633 | } | |
1634 | ||
1635 | kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu; | |
1636 | smp_wmb(); | |
1637 | atomic_inc(&kvm->online_vcpus); | |
1638 | ||
1639 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE | |
1640 | if (kvm->bsp_vcpu_id == id) | |
1641 | kvm->bsp_vcpu = vcpu; | |
1642 | #endif | |
1643 | mutex_unlock(&kvm->lock); | |
fb3f0f51 | 1644 | return r; |
39c3b86e | 1645 | |
e9b11c17 | 1646 | vcpu_destroy: |
7d8fece6 | 1647 | mutex_unlock(&kvm->lock); |
d40ccc62 | 1648 | kvm_arch_vcpu_destroy(vcpu); |
c5ea7660 AK |
1649 | return r; |
1650 | } | |
1651 | ||
1961d276 AK |
1652 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
1653 | { | |
1654 | if (sigset) { | |
1655 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
1656 | vcpu->sigset_active = 1; | |
1657 | vcpu->sigset = *sigset; | |
1658 | } else | |
1659 | vcpu->sigset_active = 0; | |
1660 | return 0; | |
1661 | } | |
1662 | ||
bccf2150 AK |
1663 | static long kvm_vcpu_ioctl(struct file *filp, |
1664 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 1665 | { |
bccf2150 | 1666 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 1667 | void __user *argp = (void __user *)arg; |
313a3dc7 | 1668 | int r; |
fa3795a7 DH |
1669 | struct kvm_fpu *fpu = NULL; |
1670 | struct kvm_sregs *kvm_sregs = NULL; | |
6aa8b732 | 1671 | |
6d4e4c4f AK |
1672 | if (vcpu->kvm->mm != current->mm) |
1673 | return -EIO; | |
2122ff5e AK |
1674 | |
1675 | #if defined(CONFIG_S390) || defined(CONFIG_PPC) | |
1676 | /* | |
1677 | * Special cases: vcpu ioctls that are asynchronous to vcpu execution, | |
1678 | * so vcpu_load() would break it. | |
1679 | */ | |
1680 | if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT) | |
1681 | return kvm_arch_vcpu_ioctl(filp, ioctl, arg); | |
1682 | #endif | |
1683 | ||
1684 | ||
1685 | vcpu_load(vcpu); | |
6aa8b732 | 1686 | switch (ioctl) { |
9a2bb7f4 | 1687 | case KVM_RUN: |
f0fe5108 AK |
1688 | r = -EINVAL; |
1689 | if (arg) | |
1690 | goto out; | |
b6c7a5dc | 1691 | r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run); |
64be5007 | 1692 | trace_kvm_userspace_exit(vcpu->run->exit_reason, r); |
6aa8b732 | 1693 | break; |
6aa8b732 | 1694 | case KVM_GET_REGS: { |
3e4bb3ac | 1695 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1696 | |
3e4bb3ac XZ |
1697 | r = -ENOMEM; |
1698 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1699 | if (!kvm_regs) | |
6aa8b732 | 1700 | goto out; |
3e4bb3ac XZ |
1701 | r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs); |
1702 | if (r) | |
1703 | goto out_free1; | |
6aa8b732 | 1704 | r = -EFAULT; |
3e4bb3ac XZ |
1705 | if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs))) |
1706 | goto out_free1; | |
6aa8b732 | 1707 | r = 0; |
3e4bb3ac XZ |
1708 | out_free1: |
1709 | kfree(kvm_regs); | |
6aa8b732 AK |
1710 | break; |
1711 | } | |
1712 | case KVM_SET_REGS: { | |
3e4bb3ac | 1713 | struct kvm_regs *kvm_regs; |
6aa8b732 | 1714 | |
3e4bb3ac XZ |
1715 | r = -ENOMEM; |
1716 | kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL); | |
1717 | if (!kvm_regs) | |
6aa8b732 | 1718 | goto out; |
3e4bb3ac XZ |
1719 | r = -EFAULT; |
1720 | if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs))) | |
1721 | goto out_free2; | |
1722 | r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs); | |
6aa8b732 | 1723 | if (r) |
3e4bb3ac | 1724 | goto out_free2; |
6aa8b732 | 1725 | r = 0; |
3e4bb3ac XZ |
1726 | out_free2: |
1727 | kfree(kvm_regs); | |
6aa8b732 AK |
1728 | break; |
1729 | } | |
1730 | case KVM_GET_SREGS: { | |
fa3795a7 DH |
1731 | kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1732 | r = -ENOMEM; | |
1733 | if (!kvm_sregs) | |
1734 | goto out; | |
1735 | r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs); | |
6aa8b732 AK |
1736 | if (r) |
1737 | goto out; | |
1738 | r = -EFAULT; | |
fa3795a7 | 1739 | if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs))) |
6aa8b732 AK |
1740 | goto out; |
1741 | r = 0; | |
1742 | break; | |
1743 | } | |
1744 | case KVM_SET_SREGS: { | |
fa3795a7 DH |
1745 | kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL); |
1746 | r = -ENOMEM; | |
1747 | if (!kvm_sregs) | |
1748 | goto out; | |
6aa8b732 | 1749 | r = -EFAULT; |
fa3795a7 | 1750 | if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs))) |
6aa8b732 | 1751 | goto out; |
fa3795a7 | 1752 | r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs); |
6aa8b732 AK |
1753 | if (r) |
1754 | goto out; | |
1755 | r = 0; | |
1756 | break; | |
1757 | } | |
62d9f0db MT |
1758 | case KVM_GET_MP_STATE: { |
1759 | struct kvm_mp_state mp_state; | |
1760 | ||
1761 | r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state); | |
1762 | if (r) | |
1763 | goto out; | |
1764 | r = -EFAULT; | |
1765 | if (copy_to_user(argp, &mp_state, sizeof mp_state)) | |
1766 | goto out; | |
1767 | r = 0; | |
1768 | break; | |
1769 | } | |
1770 | case KVM_SET_MP_STATE: { | |
1771 | struct kvm_mp_state mp_state; | |
1772 | ||
1773 | r = -EFAULT; | |
1774 | if (copy_from_user(&mp_state, argp, sizeof mp_state)) | |
1775 | goto out; | |
1776 | r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state); | |
1777 | if (r) | |
1778 | goto out; | |
1779 | r = 0; | |
1780 | break; | |
1781 | } | |
6aa8b732 AK |
1782 | case KVM_TRANSLATE: { |
1783 | struct kvm_translation tr; | |
1784 | ||
1785 | r = -EFAULT; | |
2f366987 | 1786 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 1787 | goto out; |
8b006791 | 1788 | r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
1789 | if (r) |
1790 | goto out; | |
1791 | r = -EFAULT; | |
2f366987 | 1792 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
1793 | goto out; |
1794 | r = 0; | |
1795 | break; | |
1796 | } | |
d0bfb940 JK |
1797 | case KVM_SET_GUEST_DEBUG: { |
1798 | struct kvm_guest_debug dbg; | |
6aa8b732 AK |
1799 | |
1800 | r = -EFAULT; | |
2f366987 | 1801 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 1802 | goto out; |
d0bfb940 | 1803 | r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg); |
6aa8b732 AK |
1804 | if (r) |
1805 | goto out; | |
1806 | r = 0; | |
1807 | break; | |
1808 | } | |
1961d276 AK |
1809 | case KVM_SET_SIGNAL_MASK: { |
1810 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
1811 | struct kvm_signal_mask kvm_sigmask; | |
1812 | sigset_t sigset, *p; | |
1813 | ||
1814 | p = NULL; | |
1815 | if (argp) { | |
1816 | r = -EFAULT; | |
1817 | if (copy_from_user(&kvm_sigmask, argp, | |
1818 | sizeof kvm_sigmask)) | |
1819 | goto out; | |
1820 | r = -EINVAL; | |
1821 | if (kvm_sigmask.len != sizeof sigset) | |
1822 | goto out; | |
1823 | r = -EFAULT; | |
1824 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
1825 | sizeof sigset)) | |
1826 | goto out; | |
1827 | p = &sigset; | |
1828 | } | |
376d41ff | 1829 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, p); |
1961d276 AK |
1830 | break; |
1831 | } | |
b8836737 | 1832 | case KVM_GET_FPU: { |
fa3795a7 DH |
1833 | fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1834 | r = -ENOMEM; | |
1835 | if (!fpu) | |
1836 | goto out; | |
1837 | r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu); | |
b8836737 AK |
1838 | if (r) |
1839 | goto out; | |
1840 | r = -EFAULT; | |
fa3795a7 | 1841 | if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu))) |
b8836737 AK |
1842 | goto out; |
1843 | r = 0; | |
1844 | break; | |
1845 | } | |
1846 | case KVM_SET_FPU: { | |
fa3795a7 DH |
1847 | fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL); |
1848 | r = -ENOMEM; | |
1849 | if (!fpu) | |
1850 | goto out; | |
b8836737 | 1851 | r = -EFAULT; |
fa3795a7 | 1852 | if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu))) |
b8836737 | 1853 | goto out; |
fa3795a7 | 1854 | r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu); |
b8836737 AK |
1855 | if (r) |
1856 | goto out; | |
1857 | r = 0; | |
1858 | break; | |
1859 | } | |
bccf2150 | 1860 | default: |
313a3dc7 | 1861 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
1862 | } |
1863 | out: | |
2122ff5e | 1864 | vcpu_put(vcpu); |
fa3795a7 DH |
1865 | kfree(fpu); |
1866 | kfree(kvm_sregs); | |
bccf2150 AK |
1867 | return r; |
1868 | } | |
1869 | ||
1870 | static long kvm_vm_ioctl(struct file *filp, | |
1871 | unsigned int ioctl, unsigned long arg) | |
1872 | { | |
1873 | struct kvm *kvm = filp->private_data; | |
1874 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 1875 | int r; |
bccf2150 | 1876 | |
6d4e4c4f AK |
1877 | if (kvm->mm != current->mm) |
1878 | return -EIO; | |
bccf2150 AK |
1879 | switch (ioctl) { |
1880 | case KVM_CREATE_VCPU: | |
1881 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
1882 | if (r < 0) | |
1883 | goto out; | |
1884 | break; | |
6fc138d2 IE |
1885 | case KVM_SET_USER_MEMORY_REGION: { |
1886 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
1887 | ||
1888 | r = -EFAULT; | |
1889 | if (copy_from_user(&kvm_userspace_mem, argp, | |
1890 | sizeof kvm_userspace_mem)) | |
1891 | goto out; | |
1892 | ||
1893 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
1894 | if (r) |
1895 | goto out; | |
1896 | break; | |
1897 | } | |
1898 | case KVM_GET_DIRTY_LOG: { | |
1899 | struct kvm_dirty_log log; | |
1900 | ||
1901 | r = -EFAULT; | |
2f366987 | 1902 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 1903 | goto out; |
2c6f5df9 | 1904 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
1905 | if (r) |
1906 | goto out; | |
1907 | break; | |
1908 | } | |
5f94c174 LV |
1909 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
1910 | case KVM_REGISTER_COALESCED_MMIO: { | |
1911 | struct kvm_coalesced_mmio_zone zone; | |
1912 | r = -EFAULT; | |
1913 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1914 | goto out; | |
5f94c174 LV |
1915 | r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone); |
1916 | if (r) | |
1917 | goto out; | |
1918 | r = 0; | |
1919 | break; | |
1920 | } | |
1921 | case KVM_UNREGISTER_COALESCED_MMIO: { | |
1922 | struct kvm_coalesced_mmio_zone zone; | |
1923 | r = -EFAULT; | |
1924 | if (copy_from_user(&zone, argp, sizeof zone)) | |
1925 | goto out; | |
5f94c174 LV |
1926 | r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone); |
1927 | if (r) | |
1928 | goto out; | |
1929 | r = 0; | |
1930 | break; | |
1931 | } | |
1932 | #endif | |
721eecbf GH |
1933 | case KVM_IRQFD: { |
1934 | struct kvm_irqfd data; | |
1935 | ||
1936 | r = -EFAULT; | |
1937 | if (copy_from_user(&data, argp, sizeof data)) | |
1938 | goto out; | |
1939 | r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags); | |
1940 | break; | |
1941 | } | |
d34e6b17 GH |
1942 | case KVM_IOEVENTFD: { |
1943 | struct kvm_ioeventfd data; | |
1944 | ||
1945 | r = -EFAULT; | |
1946 | if (copy_from_user(&data, argp, sizeof data)) | |
1947 | goto out; | |
1948 | r = kvm_ioeventfd(kvm, &data); | |
1949 | break; | |
1950 | } | |
73880c80 GN |
1951 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
1952 | case KVM_SET_BOOT_CPU_ID: | |
1953 | r = 0; | |
894a9c55 | 1954 | mutex_lock(&kvm->lock); |
73880c80 GN |
1955 | if (atomic_read(&kvm->online_vcpus) != 0) |
1956 | r = -EBUSY; | |
1957 | else | |
1958 | kvm->bsp_vcpu_id = arg; | |
894a9c55 | 1959 | mutex_unlock(&kvm->lock); |
73880c80 GN |
1960 | break; |
1961 | #endif | |
f17abe9a | 1962 | default: |
1fe779f8 | 1963 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
bfd99ff5 AK |
1964 | if (r == -ENOTTY) |
1965 | r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg); | |
f17abe9a AK |
1966 | } |
1967 | out: | |
1968 | return r; | |
1969 | } | |
1970 | ||
6ff5894c AB |
1971 | #ifdef CONFIG_COMPAT |
1972 | struct compat_kvm_dirty_log { | |
1973 | __u32 slot; | |
1974 | __u32 padding1; | |
1975 | union { | |
1976 | compat_uptr_t dirty_bitmap; /* one bit per page */ | |
1977 | __u64 padding2; | |
1978 | }; | |
1979 | }; | |
1980 | ||
1981 | static long kvm_vm_compat_ioctl(struct file *filp, | |
1982 | unsigned int ioctl, unsigned long arg) | |
1983 | { | |
1984 | struct kvm *kvm = filp->private_data; | |
1985 | int r; | |
1986 | ||
1987 | if (kvm->mm != current->mm) | |
1988 | return -EIO; | |
1989 | switch (ioctl) { | |
1990 | case KVM_GET_DIRTY_LOG: { | |
1991 | struct compat_kvm_dirty_log compat_log; | |
1992 | struct kvm_dirty_log log; | |
1993 | ||
1994 | r = -EFAULT; | |
1995 | if (copy_from_user(&compat_log, (void __user *)arg, | |
1996 | sizeof(compat_log))) | |
1997 | goto out; | |
1998 | log.slot = compat_log.slot; | |
1999 | log.padding1 = compat_log.padding1; | |
2000 | log.padding2 = compat_log.padding2; | |
2001 | log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap); | |
2002 | ||
2003 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); | |
2004 | if (r) | |
2005 | goto out; | |
2006 | break; | |
2007 | } | |
2008 | default: | |
2009 | r = kvm_vm_ioctl(filp, ioctl, arg); | |
2010 | } | |
2011 | ||
2012 | out: | |
2013 | return r; | |
2014 | } | |
2015 | #endif | |
2016 | ||
e4a533a4 | 2017 | static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf) |
f17abe9a | 2018 | { |
777b3f49 MT |
2019 | struct page *page[1]; |
2020 | unsigned long addr; | |
2021 | int npages; | |
2022 | gfn_t gfn = vmf->pgoff; | |
f17abe9a | 2023 | struct kvm *kvm = vma->vm_file->private_data; |
f17abe9a | 2024 | |
777b3f49 MT |
2025 | addr = gfn_to_hva(kvm, gfn); |
2026 | if (kvm_is_error_hva(addr)) | |
e4a533a4 | 2027 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2028 | |
2029 | npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page, | |
2030 | NULL); | |
2031 | if (unlikely(npages != 1)) | |
e4a533a4 | 2032 | return VM_FAULT_SIGBUS; |
777b3f49 MT |
2033 | |
2034 | vmf->page = page[0]; | |
e4a533a4 | 2035 | return 0; |
f17abe9a AK |
2036 | } |
2037 | ||
f0f37e2f | 2038 | static const struct vm_operations_struct kvm_vm_vm_ops = { |
e4a533a4 | 2039 | .fault = kvm_vm_fault, |
f17abe9a AK |
2040 | }; |
2041 | ||
2042 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2043 | { | |
2044 | vma->vm_ops = &kvm_vm_vm_ops; | |
2045 | return 0; | |
2046 | } | |
2047 | ||
3d3aab1b | 2048 | static struct file_operations kvm_vm_fops = { |
f17abe9a AK |
2049 | .release = kvm_vm_release, |
2050 | .unlocked_ioctl = kvm_vm_ioctl, | |
6ff5894c AB |
2051 | #ifdef CONFIG_COMPAT |
2052 | .compat_ioctl = kvm_vm_compat_ioctl, | |
2053 | #endif | |
f17abe9a | 2054 | .mmap = kvm_vm_mmap, |
6038f373 | 2055 | .llseek = noop_llseek, |
f17abe9a AK |
2056 | }; |
2057 | ||
2058 | static int kvm_dev_ioctl_create_vm(void) | |
2059 | { | |
aac87636 | 2060 | int r; |
f17abe9a AK |
2061 | struct kvm *kvm; |
2062 | ||
f17abe9a | 2063 | kvm = kvm_create_vm(); |
d6d28168 AK |
2064 | if (IS_ERR(kvm)) |
2065 | return PTR_ERR(kvm); | |
6ce5a090 TY |
2066 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET |
2067 | r = kvm_coalesced_mmio_init(kvm); | |
2068 | if (r < 0) { | |
2069 | kvm_put_kvm(kvm); | |
2070 | return r; | |
2071 | } | |
2072 | #endif | |
aac87636 HC |
2073 | r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR); |
2074 | if (r < 0) | |
66c0b394 | 2075 | kvm_put_kvm(kvm); |
f17abe9a | 2076 | |
aac87636 | 2077 | return r; |
f17abe9a AK |
2078 | } |
2079 | ||
1a811b61 AK |
2080 | static long kvm_dev_ioctl_check_extension_generic(long arg) |
2081 | { | |
2082 | switch (arg) { | |
ca9edaee | 2083 | case KVM_CAP_USER_MEMORY: |
1a811b61 | 2084 | case KVM_CAP_DESTROY_MEMORY_REGION_WORKS: |
4cd481f6 | 2085 | case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS: |
73880c80 GN |
2086 | #ifdef CONFIG_KVM_APIC_ARCHITECTURE |
2087 | case KVM_CAP_SET_BOOT_CPU_ID: | |
2088 | #endif | |
a9c7399d | 2089 | case KVM_CAP_INTERNAL_ERROR_DATA: |
1a811b61 | 2090 | return 1; |
399ec807 AK |
2091 | #ifdef CONFIG_HAVE_KVM_IRQCHIP |
2092 | case KVM_CAP_IRQ_ROUTING: | |
36463146 | 2093 | return KVM_MAX_IRQ_ROUTES; |
399ec807 | 2094 | #endif |
1a811b61 AK |
2095 | default: |
2096 | break; | |
2097 | } | |
2098 | return kvm_dev_ioctl_check_extension(arg); | |
2099 | } | |
2100 | ||
f17abe9a AK |
2101 | static long kvm_dev_ioctl(struct file *filp, |
2102 | unsigned int ioctl, unsigned long arg) | |
2103 | { | |
07c45a36 | 2104 | long r = -EINVAL; |
f17abe9a AK |
2105 | |
2106 | switch (ioctl) { | |
2107 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2108 | r = -EINVAL; |
2109 | if (arg) | |
2110 | goto out; | |
f17abe9a AK |
2111 | r = KVM_API_VERSION; |
2112 | break; | |
2113 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2114 | r = -EINVAL; |
2115 | if (arg) | |
2116 | goto out; | |
f17abe9a AK |
2117 | r = kvm_dev_ioctl_create_vm(); |
2118 | break; | |
018d00d2 | 2119 | case KVM_CHECK_EXTENSION: |
1a811b61 | 2120 | r = kvm_dev_ioctl_check_extension_generic(arg); |
5d308f45 | 2121 | break; |
07c45a36 AK |
2122 | case KVM_GET_VCPU_MMAP_SIZE: |
2123 | r = -EINVAL; | |
2124 | if (arg) | |
2125 | goto out; | |
adb1ff46 AK |
2126 | r = PAGE_SIZE; /* struct kvm_run */ |
2127 | #ifdef CONFIG_X86 | |
2128 | r += PAGE_SIZE; /* pio data page */ | |
5f94c174 LV |
2129 | #endif |
2130 | #ifdef KVM_COALESCED_MMIO_PAGE_OFFSET | |
2131 | r += PAGE_SIZE; /* coalesced mmio ring page */ | |
adb1ff46 | 2132 | #endif |
07c45a36 | 2133 | break; |
d4c9ff2d FEL |
2134 | case KVM_TRACE_ENABLE: |
2135 | case KVM_TRACE_PAUSE: | |
2136 | case KVM_TRACE_DISABLE: | |
2023a29c | 2137 | r = -EOPNOTSUPP; |
d4c9ff2d | 2138 | break; |
6aa8b732 | 2139 | default: |
043405e1 | 2140 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2141 | } |
2142 | out: | |
2143 | return r; | |
2144 | } | |
2145 | ||
6aa8b732 | 2146 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2147 | .unlocked_ioctl = kvm_dev_ioctl, |
2148 | .compat_ioctl = kvm_dev_ioctl, | |
6038f373 | 2149 | .llseek = noop_llseek, |
6aa8b732 AK |
2150 | }; |
2151 | ||
2152 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2153 | KVM_MINOR, |
6aa8b732 AK |
2154 | "kvm", |
2155 | &kvm_chardev_ops, | |
2156 | }; | |
2157 | ||
75b7127c | 2158 | static void hardware_enable_nolock(void *junk) |
1b6c0168 AK |
2159 | { |
2160 | int cpu = raw_smp_processor_id(); | |
10474ae8 | 2161 | int r; |
1b6c0168 | 2162 | |
7f59f492 | 2163 | if (cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2164 | return; |
10474ae8 | 2165 | |
7f59f492 | 2166 | cpumask_set_cpu(cpu, cpus_hardware_enabled); |
10474ae8 AG |
2167 | |
2168 | r = kvm_arch_hardware_enable(NULL); | |
2169 | ||
2170 | if (r) { | |
2171 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); | |
2172 | atomic_inc(&hardware_enable_failed); | |
2173 | printk(KERN_INFO "kvm: enabling virtualization on " | |
2174 | "CPU%d failed\n", cpu); | |
2175 | } | |
1b6c0168 AK |
2176 | } |
2177 | ||
75b7127c TY |
2178 | static void hardware_enable(void *junk) |
2179 | { | |
2180 | spin_lock(&kvm_lock); | |
2181 | hardware_enable_nolock(junk); | |
2182 | spin_unlock(&kvm_lock); | |
2183 | } | |
2184 | ||
2185 | static void hardware_disable_nolock(void *junk) | |
1b6c0168 AK |
2186 | { |
2187 | int cpu = raw_smp_processor_id(); | |
2188 | ||
7f59f492 | 2189 | if (!cpumask_test_cpu(cpu, cpus_hardware_enabled)) |
1b6c0168 | 2190 | return; |
7f59f492 | 2191 | cpumask_clear_cpu(cpu, cpus_hardware_enabled); |
e9b11c17 | 2192 | kvm_arch_hardware_disable(NULL); |
1b6c0168 AK |
2193 | } |
2194 | ||
75b7127c TY |
2195 | static void hardware_disable(void *junk) |
2196 | { | |
2197 | spin_lock(&kvm_lock); | |
2198 | hardware_disable_nolock(junk); | |
2199 | spin_unlock(&kvm_lock); | |
2200 | } | |
2201 | ||
10474ae8 AG |
2202 | static void hardware_disable_all_nolock(void) |
2203 | { | |
2204 | BUG_ON(!kvm_usage_count); | |
2205 | ||
2206 | kvm_usage_count--; | |
2207 | if (!kvm_usage_count) | |
75b7127c | 2208 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
10474ae8 AG |
2209 | } |
2210 | ||
2211 | static void hardware_disable_all(void) | |
2212 | { | |
2213 | spin_lock(&kvm_lock); | |
2214 | hardware_disable_all_nolock(); | |
2215 | spin_unlock(&kvm_lock); | |
2216 | } | |
2217 | ||
2218 | static int hardware_enable_all(void) | |
2219 | { | |
2220 | int r = 0; | |
2221 | ||
2222 | spin_lock(&kvm_lock); | |
2223 | ||
2224 | kvm_usage_count++; | |
2225 | if (kvm_usage_count == 1) { | |
2226 | atomic_set(&hardware_enable_failed, 0); | |
75b7127c | 2227 | on_each_cpu(hardware_enable_nolock, NULL, 1); |
10474ae8 AG |
2228 | |
2229 | if (atomic_read(&hardware_enable_failed)) { | |
2230 | hardware_disable_all_nolock(); | |
2231 | r = -EBUSY; | |
2232 | } | |
2233 | } | |
2234 | ||
2235 | spin_unlock(&kvm_lock); | |
2236 | ||
2237 | return r; | |
2238 | } | |
2239 | ||
774c47f1 AK |
2240 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2241 | void *v) | |
2242 | { | |
2243 | int cpu = (long)v; | |
2244 | ||
10474ae8 AG |
2245 | if (!kvm_usage_count) |
2246 | return NOTIFY_OK; | |
2247 | ||
1a6f4d7f | 2248 | val &= ~CPU_TASKS_FROZEN; |
774c47f1 | 2249 | switch (val) { |
cec9ad27 | 2250 | case CPU_DYING: |
6ec8a856 AK |
2251 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2252 | cpu); | |
2253 | hardware_disable(NULL); | |
2254 | break; | |
da908f2f | 2255 | case CPU_STARTING: |
43934a38 JK |
2256 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2257 | cpu); | |
da908f2f | 2258 | hardware_enable(NULL); |
774c47f1 AK |
2259 | break; |
2260 | } | |
2261 | return NOTIFY_OK; | |
2262 | } | |
2263 | ||
4ecac3fd | 2264 | |
b7c4145b | 2265 | asmlinkage void kvm_spurious_fault(void) |
4ecac3fd | 2266 | { |
4ecac3fd AK |
2267 | /* Fault while not rebooting. We want the trace. */ |
2268 | BUG(); | |
2269 | } | |
b7c4145b | 2270 | EXPORT_SYMBOL_GPL(kvm_spurious_fault); |
4ecac3fd | 2271 | |
9a2b85c6 | 2272 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2273 | void *v) |
9a2b85c6 | 2274 | { |
8e1c1815 SY |
2275 | /* |
2276 | * Some (well, at least mine) BIOSes hang on reboot if | |
2277 | * in vmx root mode. | |
2278 | * | |
2279 | * And Intel TXT required VMX off for all cpu when system shutdown. | |
2280 | */ | |
2281 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2282 | kvm_rebooting = true; | |
75b7127c | 2283 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
9a2b85c6 RR |
2284 | return NOTIFY_OK; |
2285 | } | |
2286 | ||
2287 | static struct notifier_block kvm_reboot_notifier = { | |
2288 | .notifier_call = kvm_reboot, | |
2289 | .priority = 0, | |
2290 | }; | |
2291 | ||
e93f8a0f | 2292 | static void kvm_io_bus_destroy(struct kvm_io_bus *bus) |
2eeb2e94 GH |
2293 | { |
2294 | int i; | |
2295 | ||
2296 | for (i = 0; i < bus->dev_count; i++) { | |
2297 | struct kvm_io_device *pos = bus->devs[i]; | |
2298 | ||
2299 | kvm_iodevice_destructor(pos); | |
2300 | } | |
e93f8a0f | 2301 | kfree(bus); |
2eeb2e94 GH |
2302 | } |
2303 | ||
bda9020e | 2304 | /* kvm_io_bus_write - called under kvm->slots_lock */ |
e93f8a0f | 2305 | int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
bda9020e | 2306 | int len, const void *val) |
2eeb2e94 GH |
2307 | { |
2308 | int i; | |
90d83dc3 LJ |
2309 | struct kvm_io_bus *bus; |
2310 | ||
2311 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); | |
bda9020e MT |
2312 | for (i = 0; i < bus->dev_count; i++) |
2313 | if (!kvm_iodevice_write(bus->devs[i], addr, len, val)) | |
2314 | return 0; | |
2315 | return -EOPNOTSUPP; | |
2316 | } | |
2eeb2e94 | 2317 | |
bda9020e | 2318 | /* kvm_io_bus_read - called under kvm->slots_lock */ |
e93f8a0f MT |
2319 | int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr, |
2320 | int len, void *val) | |
bda9020e MT |
2321 | { |
2322 | int i; | |
90d83dc3 | 2323 | struct kvm_io_bus *bus; |
e93f8a0f | 2324 | |
90d83dc3 | 2325 | bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu); |
bda9020e MT |
2326 | for (i = 0; i < bus->dev_count; i++) |
2327 | if (!kvm_iodevice_read(bus->devs[i], addr, len, val)) | |
2328 | return 0; | |
2329 | return -EOPNOTSUPP; | |
2eeb2e94 GH |
2330 | } |
2331 | ||
79fac95e | 2332 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2333 | int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2334 | struct kvm_io_device *dev) | |
6c474694 | 2335 | { |
e93f8a0f | 2336 | struct kvm_io_bus *new_bus, *bus; |
090b7aff | 2337 | |
e93f8a0f | 2338 | bus = kvm->buses[bus_idx]; |
090b7aff GH |
2339 | if (bus->dev_count > NR_IOBUS_DEVS-1) |
2340 | return -ENOSPC; | |
2eeb2e94 | 2341 | |
e93f8a0f MT |
2342 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2343 | if (!new_bus) | |
2344 | return -ENOMEM; | |
2345 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2346 | new_bus->devs[new_bus->dev_count++] = dev; | |
2347 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2348 | synchronize_srcu_expedited(&kvm->srcu); | |
2349 | kfree(bus); | |
090b7aff GH |
2350 | |
2351 | return 0; | |
2352 | } | |
2353 | ||
79fac95e | 2354 | /* Caller must hold slots_lock. */ |
e93f8a0f MT |
2355 | int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx, |
2356 | struct kvm_io_device *dev) | |
090b7aff | 2357 | { |
e93f8a0f MT |
2358 | int i, r; |
2359 | struct kvm_io_bus *new_bus, *bus; | |
090b7aff | 2360 | |
e93f8a0f MT |
2361 | new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL); |
2362 | if (!new_bus) | |
2363 | return -ENOMEM; | |
090b7aff | 2364 | |
e93f8a0f MT |
2365 | bus = kvm->buses[bus_idx]; |
2366 | memcpy(new_bus, bus, sizeof(struct kvm_io_bus)); | |
2367 | ||
2368 | r = -ENOENT; | |
2369 | for (i = 0; i < new_bus->dev_count; i++) | |
2370 | if (new_bus->devs[i] == dev) { | |
2371 | r = 0; | |
2372 | new_bus->devs[i] = new_bus->devs[--new_bus->dev_count]; | |
090b7aff GH |
2373 | break; |
2374 | } | |
e93f8a0f MT |
2375 | |
2376 | if (r) { | |
2377 | kfree(new_bus); | |
2378 | return r; | |
2379 | } | |
2380 | ||
2381 | rcu_assign_pointer(kvm->buses[bus_idx], new_bus); | |
2382 | synchronize_srcu_expedited(&kvm->srcu); | |
2383 | kfree(bus); | |
2384 | return r; | |
2eeb2e94 GH |
2385 | } |
2386 | ||
774c47f1 AK |
2387 | static struct notifier_block kvm_cpu_notifier = { |
2388 | .notifier_call = kvm_cpu_hotplug, | |
774c47f1 AK |
2389 | }; |
2390 | ||
8b88b099 | 2391 | static int vm_stat_get(void *_offset, u64 *val) |
ba1389b7 AK |
2392 | { |
2393 | unsigned offset = (long)_offset; | |
ba1389b7 AK |
2394 | struct kvm *kvm; |
2395 | ||
8b88b099 | 2396 | *val = 0; |
ba1389b7 AK |
2397 | spin_lock(&kvm_lock); |
2398 | list_for_each_entry(kvm, &vm_list, vm_list) | |
8b88b099 | 2399 | *val += *(u32 *)((void *)kvm + offset); |
ba1389b7 | 2400 | spin_unlock(&kvm_lock); |
8b88b099 | 2401 | return 0; |
ba1389b7 AK |
2402 | } |
2403 | ||
2404 | DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n"); | |
2405 | ||
8b88b099 | 2406 | static int vcpu_stat_get(void *_offset, u64 *val) |
1165f5fe AK |
2407 | { |
2408 | unsigned offset = (long)_offset; | |
1165f5fe AK |
2409 | struct kvm *kvm; |
2410 | struct kvm_vcpu *vcpu; | |
2411 | int i; | |
2412 | ||
8b88b099 | 2413 | *val = 0; |
1165f5fe AK |
2414 | spin_lock(&kvm_lock); |
2415 | list_for_each_entry(kvm, &vm_list, vm_list) | |
988a2cae GN |
2416 | kvm_for_each_vcpu(i, vcpu, kvm) |
2417 | *val += *(u32 *)((void *)vcpu + offset); | |
2418 | ||
1165f5fe | 2419 | spin_unlock(&kvm_lock); |
8b88b099 | 2420 | return 0; |
1165f5fe AK |
2421 | } |
2422 | ||
ba1389b7 AK |
2423 | DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n"); |
2424 | ||
828c0950 | 2425 | static const struct file_operations *stat_fops[] = { |
ba1389b7 AK |
2426 | [KVM_STAT_VCPU] = &vcpu_stat_fops, |
2427 | [KVM_STAT_VM] = &vm_stat_fops, | |
2428 | }; | |
1165f5fe | 2429 | |
a16b043c | 2430 | static void kvm_init_debug(void) |
6aa8b732 AK |
2431 | { |
2432 | struct kvm_stats_debugfs_item *p; | |
2433 | ||
76f7c879 | 2434 | kvm_debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2435 | for (p = debugfs_entries; p->name; ++p) |
76f7c879 | 2436 | p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir, |
1165f5fe | 2437 | (void *)(long)p->offset, |
ba1389b7 | 2438 | stat_fops[p->kind]); |
6aa8b732 AK |
2439 | } |
2440 | ||
2441 | static void kvm_exit_debug(void) | |
2442 | { | |
2443 | struct kvm_stats_debugfs_item *p; | |
2444 | ||
2445 | for (p = debugfs_entries; p->name; ++p) | |
2446 | debugfs_remove(p->dentry); | |
76f7c879 | 2447 | debugfs_remove(kvm_debugfs_dir); |
6aa8b732 AK |
2448 | } |
2449 | ||
59ae6c6b AK |
2450 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2451 | { | |
10474ae8 | 2452 | if (kvm_usage_count) |
75b7127c | 2453 | hardware_disable_nolock(NULL); |
59ae6c6b AK |
2454 | return 0; |
2455 | } | |
2456 | ||
2457 | static int kvm_resume(struct sys_device *dev) | |
2458 | { | |
ca84d1a2 ZA |
2459 | if (kvm_usage_count) { |
2460 | WARN_ON(spin_is_locked(&kvm_lock)); | |
75b7127c | 2461 | hardware_enable_nolock(NULL); |
ca84d1a2 | 2462 | } |
59ae6c6b AK |
2463 | return 0; |
2464 | } | |
2465 | ||
2466 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2467 | .name = "kvm", |
59ae6c6b AK |
2468 | .suspend = kvm_suspend, |
2469 | .resume = kvm_resume, | |
2470 | }; | |
2471 | ||
2472 | static struct sys_device kvm_sysdev = { | |
2473 | .id = 0, | |
2474 | .cls = &kvm_sysdev_class, | |
2475 | }; | |
2476 | ||
cea7bb21 | 2477 | struct page *bad_page; |
35149e21 | 2478 | pfn_t bad_pfn; |
6aa8b732 | 2479 | |
15ad7146 AK |
2480 | static inline |
2481 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2482 | { | |
2483 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2484 | } | |
2485 | ||
2486 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2487 | { | |
2488 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2489 | ||
e9b11c17 | 2490 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 AK |
2491 | } |
2492 | ||
2493 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2494 | struct task_struct *next) | |
2495 | { | |
2496 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2497 | ||
e9b11c17 | 2498 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
2499 | } |
2500 | ||
0ee75bea | 2501 | int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align, |
c16f862d | 2502 | struct module *module) |
6aa8b732 AK |
2503 | { |
2504 | int r; | |
002c7f7c | 2505 | int cpu; |
6aa8b732 | 2506 | |
f8c16bba ZX |
2507 | r = kvm_arch_init(opaque); |
2508 | if (r) | |
d2308784 | 2509 | goto out_fail; |
cb498ea2 ZX |
2510 | |
2511 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
2512 | ||
2513 | if (bad_page == NULL) { | |
2514 | r = -ENOMEM; | |
2515 | goto out; | |
2516 | } | |
2517 | ||
35149e21 AL |
2518 | bad_pfn = page_to_pfn(bad_page); |
2519 | ||
bf998156 HY |
2520 | hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2521 | ||
2522 | if (hwpoison_page == NULL) { | |
2523 | r = -ENOMEM; | |
2524 | goto out_free_0; | |
2525 | } | |
2526 | ||
2527 | hwpoison_pfn = page_to_pfn(hwpoison_page); | |
2528 | ||
edba23e5 GN |
2529 | fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
2530 | ||
2531 | if (fault_page == NULL) { | |
2532 | r = -ENOMEM; | |
2533 | goto out_free_0; | |
2534 | } | |
2535 | ||
2536 | fault_pfn = page_to_pfn(fault_page); | |
2537 | ||
8437a617 | 2538 | if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) { |
7f59f492 RR |
2539 | r = -ENOMEM; |
2540 | goto out_free_0; | |
2541 | } | |
2542 | ||
e9b11c17 | 2543 | r = kvm_arch_hardware_setup(); |
6aa8b732 | 2544 | if (r < 0) |
7f59f492 | 2545 | goto out_free_0a; |
6aa8b732 | 2546 | |
002c7f7c YS |
2547 | for_each_online_cpu(cpu) { |
2548 | smp_call_function_single(cpu, | |
e9b11c17 | 2549 | kvm_arch_check_processor_compat, |
8691e5a8 | 2550 | &r, 1); |
002c7f7c | 2551 | if (r < 0) |
d2308784 | 2552 | goto out_free_1; |
002c7f7c YS |
2553 | } |
2554 | ||
774c47f1 AK |
2555 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2556 | if (r) | |
d2308784 | 2557 | goto out_free_2; |
6aa8b732 AK |
2558 | register_reboot_notifier(&kvm_reboot_notifier); |
2559 | ||
59ae6c6b AK |
2560 | r = sysdev_class_register(&kvm_sysdev_class); |
2561 | if (r) | |
d2308784 | 2562 | goto out_free_3; |
59ae6c6b AK |
2563 | |
2564 | r = sysdev_register(&kvm_sysdev); | |
2565 | if (r) | |
d2308784 | 2566 | goto out_free_4; |
59ae6c6b | 2567 | |
c16f862d | 2568 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
0ee75bea AK |
2569 | if (!vcpu_align) |
2570 | vcpu_align = __alignof__(struct kvm_vcpu); | |
2571 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align, | |
56919c5c | 2572 | 0, NULL); |
c16f862d RR |
2573 | if (!kvm_vcpu_cache) { |
2574 | r = -ENOMEM; | |
d2308784 | 2575 | goto out_free_5; |
c16f862d RR |
2576 | } |
2577 | ||
af585b92 GN |
2578 | r = kvm_async_pf_init(); |
2579 | if (r) | |
2580 | goto out_free; | |
2581 | ||
6aa8b732 | 2582 | kvm_chardev_ops.owner = module; |
3d3aab1b CB |
2583 | kvm_vm_fops.owner = module; |
2584 | kvm_vcpu_fops.owner = module; | |
6aa8b732 AK |
2585 | |
2586 | r = misc_register(&kvm_dev); | |
2587 | if (r) { | |
d77c26fc | 2588 | printk(KERN_ERR "kvm: misc device register failed\n"); |
af585b92 | 2589 | goto out_unreg; |
6aa8b732 AK |
2590 | } |
2591 | ||
15ad7146 AK |
2592 | kvm_preempt_ops.sched_in = kvm_sched_in; |
2593 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
2594 | ||
0ea4ed8e DW |
2595 | kvm_init_debug(); |
2596 | ||
c7addb90 | 2597 | return 0; |
6aa8b732 | 2598 | |
af585b92 GN |
2599 | out_unreg: |
2600 | kvm_async_pf_deinit(); | |
6aa8b732 | 2601 | out_free: |
c16f862d | 2602 | kmem_cache_destroy(kvm_vcpu_cache); |
d2308784 | 2603 | out_free_5: |
59ae6c6b | 2604 | sysdev_unregister(&kvm_sysdev); |
d2308784 | 2605 | out_free_4: |
59ae6c6b | 2606 | sysdev_class_unregister(&kvm_sysdev_class); |
d2308784 | 2607 | out_free_3: |
6aa8b732 | 2608 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 | 2609 | unregister_cpu_notifier(&kvm_cpu_notifier); |
d2308784 | 2610 | out_free_2: |
d2308784 | 2611 | out_free_1: |
e9b11c17 | 2612 | kvm_arch_hardware_unsetup(); |
7f59f492 RR |
2613 | out_free_0a: |
2614 | free_cpumask_var(cpus_hardware_enabled); | |
d2308784 | 2615 | out_free_0: |
edba23e5 GN |
2616 | if (fault_page) |
2617 | __free_page(fault_page); | |
bf998156 HY |
2618 | if (hwpoison_page) |
2619 | __free_page(hwpoison_page); | |
d2308784 | 2620 | __free_page(bad_page); |
ca45aaae | 2621 | out: |
f8c16bba | 2622 | kvm_arch_exit(); |
d2308784 | 2623 | out_fail: |
6aa8b732 AK |
2624 | return r; |
2625 | } | |
cb498ea2 | 2626 | EXPORT_SYMBOL_GPL(kvm_init); |
6aa8b732 | 2627 | |
cb498ea2 | 2628 | void kvm_exit(void) |
6aa8b732 | 2629 | { |
0ea4ed8e | 2630 | kvm_exit_debug(); |
6aa8b732 | 2631 | misc_deregister(&kvm_dev); |
c16f862d | 2632 | kmem_cache_destroy(kvm_vcpu_cache); |
af585b92 | 2633 | kvm_async_pf_deinit(); |
59ae6c6b AK |
2634 | sysdev_unregister(&kvm_sysdev); |
2635 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 2636 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 2637 | unregister_cpu_notifier(&kvm_cpu_notifier); |
75b7127c | 2638 | on_each_cpu(hardware_disable_nolock, NULL, 1); |
e9b11c17 | 2639 | kvm_arch_hardware_unsetup(); |
f8c16bba | 2640 | kvm_arch_exit(); |
7f59f492 | 2641 | free_cpumask_var(cpus_hardware_enabled); |
bf998156 | 2642 | __free_page(hwpoison_page); |
cea7bb21 | 2643 | __free_page(bad_page); |
6aa8b732 | 2644 | } |
cb498ea2 | 2645 | EXPORT_SYMBOL_GPL(kvm_exit); |