]>
Commit | Line | Data |
---|---|---|
6aa8b732 AK |
1 | /* |
2 | * Kernel-based Virtual Machine driver for Linux | |
3 | * | |
4 | * This module enables machines with Intel VT-x extensions to run virtual | |
5 | * machines without emulation or binary translation. | |
6 | * | |
7 | * Copyright (C) 2006 Qumranet, Inc. | |
8 | * | |
9 | * Authors: | |
10 | * Avi Kivity <avi@qumranet.com> | |
11 | * Yaniv Kamay <yaniv@qumranet.com> | |
12 | * | |
13 | * This work is licensed under the terms of the GNU GPL, version 2. See | |
14 | * the COPYING file in the top-level directory. | |
15 | * | |
16 | */ | |
17 | ||
18 | #include "kvm.h" | |
043405e1 | 19 | #include "x86.h" |
e495606d | 20 | #include "x86_emulate.h" |
85f455f7 | 21 | #include "irq.h" |
6aa8b732 AK |
22 | |
23 | #include <linux/kvm.h> | |
24 | #include <linux/module.h> | |
25 | #include <linux/errno.h> | |
6aa8b732 AK |
26 | #include <linux/percpu.h> |
27 | #include <linux/gfp.h> | |
6aa8b732 AK |
28 | #include <linux/mm.h> |
29 | #include <linux/miscdevice.h> | |
30 | #include <linux/vmalloc.h> | |
6aa8b732 | 31 | #include <linux/reboot.h> |
6aa8b732 AK |
32 | #include <linux/debugfs.h> |
33 | #include <linux/highmem.h> | |
34 | #include <linux/file.h> | |
59ae6c6b | 35 | #include <linux/sysdev.h> |
774c47f1 | 36 | #include <linux/cpu.h> |
e8edc6e0 | 37 | #include <linux/sched.h> |
d9e368d6 AK |
38 | #include <linux/cpumask.h> |
39 | #include <linux/smp.h> | |
d6d28168 | 40 | #include <linux/anon_inodes.h> |
04d2cc77 | 41 | #include <linux/profile.h> |
7aa81cc0 | 42 | #include <linux/kvm_para.h> |
6fc138d2 | 43 | #include <linux/pagemap.h> |
8d4e1288 | 44 | #include <linux/mman.h> |
6aa8b732 | 45 | |
e495606d AK |
46 | #include <asm/processor.h> |
47 | #include <asm/msr.h> | |
48 | #include <asm/io.h> | |
49 | #include <asm/uaccess.h> | |
50 | #include <asm/desc.h> | |
6aa8b732 AK |
51 | |
52 | MODULE_AUTHOR("Qumranet"); | |
53 | MODULE_LICENSE("GPL"); | |
54 | ||
133de902 AK |
55 | static DEFINE_SPINLOCK(kvm_lock); |
56 | static LIST_HEAD(vm_list); | |
57 | ||
1b6c0168 AK |
58 | static cpumask_t cpus_hardware_enabled; |
59 | ||
cbdd1bea | 60 | struct kvm_x86_ops *kvm_x86_ops; |
c16f862d RR |
61 | struct kmem_cache *kvm_vcpu_cache; |
62 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 63 | |
15ad7146 AK |
64 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
65 | ||
1165f5fe | 66 | #define STAT_OFFSET(x) offsetof(struct kvm_vcpu, stat.x) |
6aa8b732 AK |
67 | |
68 | static struct kvm_stats_debugfs_item { | |
69 | const char *name; | |
1165f5fe | 70 | int offset; |
6aa8b732 AK |
71 | struct dentry *dentry; |
72 | } debugfs_entries[] = { | |
1165f5fe AK |
73 | { "pf_fixed", STAT_OFFSET(pf_fixed) }, |
74 | { "pf_guest", STAT_OFFSET(pf_guest) }, | |
75 | { "tlb_flush", STAT_OFFSET(tlb_flush) }, | |
76 | { "invlpg", STAT_OFFSET(invlpg) }, | |
77 | { "exits", STAT_OFFSET(exits) }, | |
78 | { "io_exits", STAT_OFFSET(io_exits) }, | |
79 | { "mmio_exits", STAT_OFFSET(mmio_exits) }, | |
80 | { "signal_exits", STAT_OFFSET(signal_exits) }, | |
81 | { "irq_window", STAT_OFFSET(irq_window_exits) }, | |
82 | { "halt_exits", STAT_OFFSET(halt_exits) }, | |
b6958ce4 | 83 | { "halt_wakeup", STAT_OFFSET(halt_wakeup) }, |
1165f5fe AK |
84 | { "request_irq", STAT_OFFSET(request_irq_exits) }, |
85 | { "irq_exits", STAT_OFFSET(irq_exits) }, | |
e6adf283 | 86 | { "light_exits", STAT_OFFSET(light_exits) }, |
2cc51560 | 87 | { "efer_reload", STAT_OFFSET(efer_reload) }, |
1165f5fe | 88 | { NULL } |
6aa8b732 AK |
89 | }; |
90 | ||
91 | static struct dentry *debugfs_dir; | |
92 | ||
6aa8b732 AK |
93 | #define EFER_RESERVED_BITS 0xfffffffffffff2fe |
94 | ||
bccf2150 AK |
95 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
96 | unsigned long arg); | |
97 | ||
5aacf0ca JM |
98 | static inline int valid_vcpu(int n) |
99 | { | |
100 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
101 | } | |
102 | ||
7702fd1f AK |
103 | void kvm_load_guest_fpu(struct kvm_vcpu *vcpu) |
104 | { | |
105 | if (!vcpu->fpu_active || vcpu->guest_fpu_loaded) | |
106 | return; | |
107 | ||
108 | vcpu->guest_fpu_loaded = 1; | |
b114b080 RR |
109 | fx_save(&vcpu->host_fx_image); |
110 | fx_restore(&vcpu->guest_fx_image); | |
7702fd1f AK |
111 | } |
112 | EXPORT_SYMBOL_GPL(kvm_load_guest_fpu); | |
113 | ||
114 | void kvm_put_guest_fpu(struct kvm_vcpu *vcpu) | |
115 | { | |
116 | if (!vcpu->guest_fpu_loaded) | |
117 | return; | |
118 | ||
119 | vcpu->guest_fpu_loaded = 0; | |
b114b080 RR |
120 | fx_save(&vcpu->guest_fx_image); |
121 | fx_restore(&vcpu->host_fx_image); | |
7702fd1f AK |
122 | } |
123 | EXPORT_SYMBOL_GPL(kvm_put_guest_fpu); | |
124 | ||
bccf2150 AK |
125 | /* |
126 | * Switches to specified vcpu, until a matching vcpu_put() | |
127 | */ | |
313a3dc7 | 128 | void vcpu_load(struct kvm_vcpu *vcpu) |
6aa8b732 | 129 | { |
15ad7146 AK |
130 | int cpu; |
131 | ||
bccf2150 | 132 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
133 | cpu = get_cpu(); |
134 | preempt_notifier_register(&vcpu->preempt_notifier); | |
313a3dc7 | 135 | kvm_arch_vcpu_load(vcpu, cpu); |
15ad7146 | 136 | put_cpu(); |
6aa8b732 AK |
137 | } |
138 | ||
313a3dc7 | 139 | void vcpu_put(struct kvm_vcpu *vcpu) |
6aa8b732 | 140 | { |
15ad7146 | 141 | preempt_disable(); |
313a3dc7 | 142 | kvm_arch_vcpu_put(vcpu); |
15ad7146 AK |
143 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
144 | preempt_enable(); | |
6aa8b732 AK |
145 | mutex_unlock(&vcpu->mutex); |
146 | } | |
147 | ||
d9e368d6 AK |
148 | static void ack_flush(void *_completed) |
149 | { | |
d9e368d6 AK |
150 | } |
151 | ||
152 | void kvm_flush_remote_tlbs(struct kvm *kvm) | |
153 | { | |
49d3bd7e | 154 | int i, cpu; |
d9e368d6 AK |
155 | cpumask_t cpus; |
156 | struct kvm_vcpu *vcpu; | |
d9e368d6 | 157 | |
d9e368d6 | 158 | cpus_clear(cpus); |
fb3f0f51 RR |
159 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
160 | vcpu = kvm->vcpus[i]; | |
161 | if (!vcpu) | |
162 | continue; | |
3176bc3e | 163 | if (test_and_set_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests)) |
d9e368d6 AK |
164 | continue; |
165 | cpu = vcpu->cpu; | |
166 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
49d3bd7e | 167 | cpu_set(cpu, cpus); |
d9e368d6 | 168 | } |
49d3bd7e | 169 | smp_call_function_mask(cpus, ack_flush, NULL, 1); |
d9e368d6 AK |
170 | } |
171 | ||
fb3f0f51 RR |
172 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
173 | { | |
174 | struct page *page; | |
175 | int r; | |
176 | ||
177 | mutex_init(&vcpu->mutex); | |
178 | vcpu->cpu = -1; | |
179 | vcpu->mmu.root_hpa = INVALID_PAGE; | |
180 | vcpu->kvm = kvm; | |
181 | vcpu->vcpu_id = id; | |
c5ec1534 HQ |
182 | if (!irqchip_in_kernel(kvm) || id == 0) |
183 | vcpu->mp_state = VCPU_MP_STATE_RUNNABLE; | |
184 | else | |
185 | vcpu->mp_state = VCPU_MP_STATE_UNINITIALIZED; | |
b6958ce4 | 186 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
187 | |
188 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
189 | if (!page) { | |
190 | r = -ENOMEM; | |
191 | goto fail; | |
192 | } | |
193 | vcpu->run = page_address(page); | |
194 | ||
195 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
196 | if (!page) { | |
197 | r = -ENOMEM; | |
198 | goto fail_free_run; | |
199 | } | |
200 | vcpu->pio_data = page_address(page); | |
201 | ||
fb3f0f51 RR |
202 | r = kvm_mmu_create(vcpu); |
203 | if (r < 0) | |
204 | goto fail_free_pio_data; | |
205 | ||
76fafa5e RR |
206 | if (irqchip_in_kernel(kvm)) { |
207 | r = kvm_create_lapic(vcpu); | |
208 | if (r < 0) | |
209 | goto fail_mmu_destroy; | |
210 | } | |
211 | ||
fb3f0f51 RR |
212 | return 0; |
213 | ||
76fafa5e RR |
214 | fail_mmu_destroy: |
215 | kvm_mmu_destroy(vcpu); | |
fb3f0f51 RR |
216 | fail_free_pio_data: |
217 | free_page((unsigned long)vcpu->pio_data); | |
218 | fail_free_run: | |
219 | free_page((unsigned long)vcpu->run); | |
220 | fail: | |
76fafa5e | 221 | return r; |
fb3f0f51 RR |
222 | } |
223 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
224 | ||
225 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
226 | { | |
d589444e | 227 | kvm_free_lapic(vcpu); |
fb3f0f51 RR |
228 | kvm_mmu_destroy(vcpu); |
229 | free_page((unsigned long)vcpu->pio_data); | |
230 | free_page((unsigned long)vcpu->run); | |
231 | } | |
232 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
233 | ||
f17abe9a | 234 | static struct kvm *kvm_create_vm(void) |
6aa8b732 AK |
235 | { |
236 | struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL); | |
6aa8b732 AK |
237 | |
238 | if (!kvm) | |
f17abe9a | 239 | return ERR_PTR(-ENOMEM); |
6aa8b732 | 240 | |
74906345 | 241 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 242 | mutex_init(&kvm->lock); |
6aa8b732 | 243 | INIT_LIST_HEAD(&kvm->active_mmu_pages); |
2eeb2e94 | 244 | kvm_io_bus_init(&kvm->mmio_bus); |
5e58cfe4 RR |
245 | spin_lock(&kvm_lock); |
246 | list_add(&kvm->vm_list, &vm_list); | |
247 | spin_unlock(&kvm_lock); | |
f17abe9a AK |
248 | return kvm; |
249 | } | |
250 | ||
6aa8b732 AK |
251 | /* |
252 | * Free any memory in @free but not in @dont. | |
253 | */ | |
254 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
255 | struct kvm_memory_slot *dont) | |
256 | { | |
290fc38d IE |
257 | if (!dont || free->rmap != dont->rmap) |
258 | vfree(free->rmap); | |
6aa8b732 AK |
259 | |
260 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
261 | vfree(free->dirty_bitmap); | |
262 | ||
6aa8b732 | 263 | free->npages = 0; |
8b6d44c7 | 264 | free->dirty_bitmap = NULL; |
8d4e1288 | 265 | free->rmap = NULL; |
6aa8b732 AK |
266 | } |
267 | ||
268 | static void kvm_free_physmem(struct kvm *kvm) | |
269 | { | |
270 | int i; | |
271 | ||
272 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 273 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
274 | } |
275 | ||
039576c0 AK |
276 | static void free_pio_guest_pages(struct kvm_vcpu *vcpu) |
277 | { | |
278 | int i; | |
279 | ||
3077c451 | 280 | for (i = 0; i < ARRAY_SIZE(vcpu->pio.guest_pages); ++i) |
039576c0 | 281 | if (vcpu->pio.guest_pages[i]) { |
8a7ae055 | 282 | kvm_release_page(vcpu->pio.guest_pages[i]); |
039576c0 AK |
283 | vcpu->pio.guest_pages[i] = NULL; |
284 | } | |
285 | } | |
286 | ||
7b53aa56 AK |
287 | static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu) |
288 | { | |
7b53aa56 AK |
289 | vcpu_load(vcpu); |
290 | kvm_mmu_unload(vcpu); | |
291 | vcpu_put(vcpu); | |
292 | } | |
293 | ||
6aa8b732 AK |
294 | static void kvm_free_vcpus(struct kvm *kvm) |
295 | { | |
296 | unsigned int i; | |
297 | ||
7b53aa56 AK |
298 | /* |
299 | * Unpin any mmu pages first. | |
300 | */ | |
301 | for (i = 0; i < KVM_MAX_VCPUS; ++i) | |
fb3f0f51 RR |
302 | if (kvm->vcpus[i]) |
303 | kvm_unload_vcpu_mmu(kvm->vcpus[i]); | |
304 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
305 | if (kvm->vcpus[i]) { | |
cbdd1bea | 306 | kvm_x86_ops->vcpu_free(kvm->vcpus[i]); |
fb3f0f51 RR |
307 | kvm->vcpus[i] = NULL; |
308 | } | |
309 | } | |
310 | ||
6aa8b732 AK |
311 | } |
312 | ||
f17abe9a AK |
313 | static void kvm_destroy_vm(struct kvm *kvm) |
314 | { | |
133de902 AK |
315 | spin_lock(&kvm_lock); |
316 | list_del(&kvm->vm_list); | |
317 | spin_unlock(&kvm_lock); | |
74906345 | 318 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 319 | kvm_io_bus_destroy(&kvm->mmio_bus); |
85f455f7 | 320 | kfree(kvm->vpic); |
1fd4f2a5 | 321 | kfree(kvm->vioapic); |
6aa8b732 AK |
322 | kvm_free_vcpus(kvm); |
323 | kvm_free_physmem(kvm); | |
324 | kfree(kvm); | |
f17abe9a AK |
325 | } |
326 | ||
327 | static int kvm_vm_release(struct inode *inode, struct file *filp) | |
328 | { | |
329 | struct kvm *kvm = filp->private_data; | |
330 | ||
331 | kvm_destroy_vm(kvm); | |
6aa8b732 AK |
332 | return 0; |
333 | } | |
334 | ||
335 | static void inject_gp(struct kvm_vcpu *vcpu) | |
336 | { | |
cbdd1bea | 337 | kvm_x86_ops->inject_gp(vcpu, 0); |
6aa8b732 AK |
338 | } |
339 | ||
6aa8b732 AK |
340 | void fx_init(struct kvm_vcpu *vcpu) |
341 | { | |
b114b080 | 342 | unsigned after_mxcsr_mask; |
6aa8b732 | 343 | |
9bd01506 RR |
344 | /* Initialize guest FPU by resetting ours and saving into guest's */ |
345 | preempt_disable(); | |
b114b080 | 346 | fx_save(&vcpu->host_fx_image); |
6aa8b732 | 347 | fpu_init(); |
b114b080 RR |
348 | fx_save(&vcpu->guest_fx_image); |
349 | fx_restore(&vcpu->host_fx_image); | |
9bd01506 | 350 | preempt_enable(); |
6aa8b732 | 351 | |
380102c8 | 352 | vcpu->cr0 |= X86_CR0_ET; |
b114b080 RR |
353 | after_mxcsr_mask = offsetof(struct i387_fxsave_struct, st_space); |
354 | vcpu->guest_fx_image.mxcsr = 0x1f80; | |
355 | memset((void *)&vcpu->guest_fx_image + after_mxcsr_mask, | |
356 | 0, sizeof(struct i387_fxsave_struct) - after_mxcsr_mask); | |
6aa8b732 AK |
357 | } |
358 | EXPORT_SYMBOL_GPL(fx_init); | |
359 | ||
6aa8b732 AK |
360 | /* |
361 | * Allocate some memory and give it an address in the guest physical address | |
362 | * space. | |
363 | * | |
364 | * Discontiguous memory is allowed, mostly for framebuffers. | |
365 | */ | |
210c7c4d IE |
366 | int kvm_set_memory_region(struct kvm *kvm, |
367 | struct kvm_userspace_memory_region *mem, | |
368 | int user_alloc) | |
6aa8b732 AK |
369 | { |
370 | int r; | |
371 | gfn_t base_gfn; | |
372 | unsigned long npages; | |
373 | unsigned long i; | |
374 | struct kvm_memory_slot *memslot; | |
375 | struct kvm_memory_slot old, new; | |
6aa8b732 AK |
376 | |
377 | r = -EINVAL; | |
378 | /* General sanity checks */ | |
379 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
380 | goto out; | |
381 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
382 | goto out; | |
e0d62c7f | 383 | if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS) |
6aa8b732 AK |
384 | goto out; |
385 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
386 | goto out; | |
387 | ||
388 | memslot = &kvm->memslots[mem->slot]; | |
389 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
390 | npages = mem->memory_size >> PAGE_SHIFT; | |
391 | ||
392 | if (!npages) | |
393 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
394 | ||
11ec2804 | 395 | mutex_lock(&kvm->lock); |
6aa8b732 | 396 | |
6aa8b732 AK |
397 | new = old = *memslot; |
398 | ||
399 | new.base_gfn = base_gfn; | |
400 | new.npages = npages; | |
401 | new.flags = mem->flags; | |
402 | ||
403 | /* Disallow changing a memory slot's size. */ | |
404 | r = -EINVAL; | |
405 | if (npages && old.npages && npages != old.npages) | |
406 | goto out_unlock; | |
407 | ||
408 | /* Check for overlaps */ | |
409 | r = -EEXIST; | |
410 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
411 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
412 | ||
413 | if (s == memslot) | |
414 | continue; | |
415 | if (!((base_gfn + npages <= s->base_gfn) || | |
416 | (base_gfn >= s->base_gfn + s->npages))) | |
417 | goto out_unlock; | |
418 | } | |
6aa8b732 | 419 | |
6aa8b732 AK |
420 | /* Free page dirty bitmap if unneeded */ |
421 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 422 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
423 | |
424 | r = -ENOMEM; | |
425 | ||
426 | /* Allocate if a slot is being created */ | |
8d4e1288 | 427 | if (npages && !new.rmap) { |
d77c26fc | 428 | new.rmap = vmalloc(npages * sizeof(struct page *)); |
290fc38d IE |
429 | |
430 | if (!new.rmap) | |
431 | goto out_unlock; | |
432 | ||
290fc38d | 433 | memset(new.rmap, 0, npages * sizeof(*new.rmap)); |
8d4e1288 | 434 | |
80b14b5b | 435 | new.user_alloc = user_alloc; |
8d4e1288 | 436 | if (user_alloc) |
8a7ae055 | 437 | new.userspace_addr = mem->userspace_addr; |
8d4e1288 AL |
438 | else { |
439 | down_write(¤t->mm->mmap_sem); | |
440 | new.userspace_addr = do_mmap(NULL, 0, | |
441 | npages * PAGE_SIZE, | |
442 | PROT_READ | PROT_WRITE, | |
443 | MAP_SHARED | MAP_ANONYMOUS, | |
444 | 0); | |
445 | up_write(¤t->mm->mmap_sem); | |
446 | ||
447 | if (IS_ERR((void *)new.userspace_addr)) | |
448 | goto out_unlock; | |
6aa8b732 | 449 | } |
80b14b5b IE |
450 | } else { |
451 | if (!old.user_alloc && old.rmap) { | |
452 | int ret; | |
453 | ||
454 | down_write(¤t->mm->mmap_sem); | |
455 | ret = do_munmap(current->mm, old.userspace_addr, | |
456 | old.npages * PAGE_SIZE); | |
457 | up_write(¤t->mm->mmap_sem); | |
458 | if (ret < 0) | |
459 | printk(KERN_WARNING | |
460 | "kvm_vm_ioctl_set_memory_region: " | |
461 | "failed to munmap memory\n"); | |
462 | } | |
6aa8b732 AK |
463 | } |
464 | ||
465 | /* Allocate page dirty bitmap if needed */ | |
466 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
467 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
468 | ||
469 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
470 | if (!new.dirty_bitmap) | |
0d8d2bd4 | 471 | goto out_unlock; |
6aa8b732 AK |
472 | memset(new.dirty_bitmap, 0, dirty_bytes); |
473 | } | |
474 | ||
6aa8b732 AK |
475 | if (mem->slot >= kvm->nmemslots) |
476 | kvm->nmemslots = mem->slot + 1; | |
477 | ||
82ce2c96 IE |
478 | if (!kvm->n_requested_mmu_pages) { |
479 | unsigned int n_pages; | |
480 | ||
481 | if (npages) { | |
482 | n_pages = npages * KVM_PERMILLE_MMU_PAGES / 1000; | |
483 | kvm_mmu_change_mmu_pages(kvm, kvm->n_alloc_mmu_pages + | |
484 | n_pages); | |
485 | } else { | |
486 | unsigned int nr_mmu_pages; | |
487 | ||
488 | n_pages = old.npages * KVM_PERMILLE_MMU_PAGES / 1000; | |
489 | nr_mmu_pages = kvm->n_alloc_mmu_pages - n_pages; | |
490 | nr_mmu_pages = max(nr_mmu_pages, | |
491 | (unsigned int) KVM_MIN_ALLOC_MMU_PAGES); | |
492 | kvm_mmu_change_mmu_pages(kvm, nr_mmu_pages); | |
493 | } | |
494 | } | |
495 | ||
6aa8b732 | 496 | *memslot = new; |
6aa8b732 | 497 | |
90cb0529 AK |
498 | kvm_mmu_slot_remove_write_access(kvm, mem->slot); |
499 | kvm_flush_remote_tlbs(kvm); | |
6aa8b732 | 500 | |
11ec2804 | 501 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
502 | |
503 | kvm_free_physmem_slot(&old, &new); | |
504 | return 0; | |
505 | ||
506 | out_unlock: | |
11ec2804 | 507 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
508 | kvm_free_physmem_slot(&new, &old); |
509 | out: | |
510 | return r; | |
210c7c4d IE |
511 | |
512 | } | |
513 | EXPORT_SYMBOL_GPL(kvm_set_memory_region); | |
514 | ||
1fe779f8 CO |
515 | int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
516 | struct | |
517 | kvm_userspace_memory_region *mem, | |
518 | int user_alloc) | |
210c7c4d | 519 | { |
e0d62c7f IE |
520 | if (mem->slot >= KVM_MEMORY_SLOTS) |
521 | return -EINVAL; | |
210c7c4d | 522 | return kvm_set_memory_region(kvm, mem, user_alloc); |
6aa8b732 AK |
523 | } |
524 | ||
525 | /* | |
526 | * Get (and clear) the dirty memory log for a memory slot. | |
527 | */ | |
2c6f5df9 AK |
528 | static int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, |
529 | struct kvm_dirty_log *log) | |
6aa8b732 AK |
530 | { |
531 | struct kvm_memory_slot *memslot; | |
532 | int r, i; | |
533 | int n; | |
534 | unsigned long any = 0; | |
535 | ||
11ec2804 | 536 | mutex_lock(&kvm->lock); |
6aa8b732 | 537 | |
6aa8b732 AK |
538 | r = -EINVAL; |
539 | if (log->slot >= KVM_MEMORY_SLOTS) | |
540 | goto out; | |
541 | ||
542 | memslot = &kvm->memslots[log->slot]; | |
543 | r = -ENOENT; | |
544 | if (!memslot->dirty_bitmap) | |
545 | goto out; | |
546 | ||
cd1a4a98 | 547 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 548 | |
cd1a4a98 | 549 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
550 | any = memslot->dirty_bitmap[i]; |
551 | ||
552 | r = -EFAULT; | |
553 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
554 | goto out; | |
555 | ||
39214915 RR |
556 | /* If nothing is dirty, don't bother messing with page tables. */ |
557 | if (any) { | |
39214915 RR |
558 | kvm_mmu_slot_remove_write_access(kvm, log->slot); |
559 | kvm_flush_remote_tlbs(kvm); | |
560 | memset(memslot->dirty_bitmap, 0, n); | |
39214915 | 561 | } |
6aa8b732 AK |
562 | |
563 | r = 0; | |
564 | ||
565 | out: | |
11ec2804 | 566 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
567 | return r; |
568 | } | |
569 | ||
cea7bb21 IE |
570 | int is_error_page(struct page *page) |
571 | { | |
572 | return page == bad_page; | |
573 | } | |
574 | EXPORT_SYMBOL_GPL(is_error_page); | |
575 | ||
290fc38d | 576 | gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn) |
e8207547 AK |
577 | { |
578 | int i; | |
579 | struct kvm_mem_alias *alias; | |
580 | ||
581 | for (i = 0; i < kvm->naliases; ++i) { | |
582 | alias = &kvm->aliases[i]; | |
583 | if (gfn >= alias->base_gfn | |
584 | && gfn < alias->base_gfn + alias->npages) | |
585 | return alias->target_gfn + gfn - alias->base_gfn; | |
586 | } | |
587 | return gfn; | |
588 | } | |
589 | ||
590 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
6aa8b732 AK |
591 | { |
592 | int i; | |
593 | ||
594 | for (i = 0; i < kvm->nmemslots; ++i) { | |
595 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
596 | ||
597 | if (gfn >= memslot->base_gfn | |
598 | && gfn < memslot->base_gfn + memslot->npages) | |
599 | return memslot; | |
600 | } | |
8b6d44c7 | 601 | return NULL; |
6aa8b732 | 602 | } |
e8207547 AK |
603 | |
604 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
605 | { | |
606 | gfn = unalias_gfn(kvm, gfn); | |
607 | return __gfn_to_memslot(kvm, gfn); | |
608 | } | |
6aa8b732 | 609 | |
e0d62c7f IE |
610 | int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn) |
611 | { | |
612 | int i; | |
613 | ||
614 | gfn = unalias_gfn(kvm, gfn); | |
615 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
616 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
617 | ||
618 | if (gfn >= memslot->base_gfn | |
619 | && gfn < memslot->base_gfn + memslot->npages) | |
620 | return 1; | |
621 | } | |
622 | return 0; | |
623 | } | |
624 | EXPORT_SYMBOL_GPL(kvm_is_visible_gfn); | |
625 | ||
954bbbc2 AK |
626 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
627 | { | |
628 | struct kvm_memory_slot *slot; | |
8d4e1288 AL |
629 | struct page *page[1]; |
630 | int npages; | |
954bbbc2 | 631 | |
60395224 AK |
632 | might_sleep(); |
633 | ||
e8207547 AK |
634 | gfn = unalias_gfn(kvm, gfn); |
635 | slot = __gfn_to_memslot(kvm, gfn); | |
8a7ae055 IE |
636 | if (!slot) { |
637 | get_page(bad_page); | |
cea7bb21 | 638 | return bad_page; |
8a7ae055 | 639 | } |
8d4e1288 AL |
640 | |
641 | down_read(¤t->mm->mmap_sem); | |
642 | npages = get_user_pages(current, current->mm, | |
643 | slot->userspace_addr | |
644 | + (gfn - slot->base_gfn) * PAGE_SIZE, 1, | |
645 | 1, 1, page, NULL); | |
646 | up_read(¤t->mm->mmap_sem); | |
647 | if (npages != 1) { | |
648 | get_page(bad_page); | |
649 | return bad_page; | |
8a7ae055 | 650 | } |
8d4e1288 AL |
651 | |
652 | return page[0]; | |
954bbbc2 AK |
653 | } |
654 | EXPORT_SYMBOL_GPL(gfn_to_page); | |
655 | ||
8a7ae055 IE |
656 | void kvm_release_page(struct page *page) |
657 | { | |
658 | if (!PageReserved(page)) | |
659 | SetPageDirty(page); | |
660 | put_page(page); | |
661 | } | |
662 | EXPORT_SYMBOL_GPL(kvm_release_page); | |
663 | ||
195aefde IE |
664 | static int next_segment(unsigned long len, int offset) |
665 | { | |
666 | if (len > PAGE_SIZE - offset) | |
667 | return PAGE_SIZE - offset; | |
668 | else | |
669 | return len; | |
670 | } | |
671 | ||
672 | int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset, | |
673 | int len) | |
674 | { | |
675 | void *page_virt; | |
676 | struct page *page; | |
677 | ||
678 | page = gfn_to_page(kvm, gfn); | |
8a7ae055 IE |
679 | if (is_error_page(page)) { |
680 | kvm_release_page(page); | |
195aefde | 681 | return -EFAULT; |
8a7ae055 | 682 | } |
195aefde IE |
683 | page_virt = kmap_atomic(page, KM_USER0); |
684 | ||
685 | memcpy(data, page_virt + offset, len); | |
686 | ||
687 | kunmap_atomic(page_virt, KM_USER0); | |
8a7ae055 | 688 | kvm_release_page(page); |
195aefde IE |
689 | return 0; |
690 | } | |
691 | EXPORT_SYMBOL_GPL(kvm_read_guest_page); | |
692 | ||
693 | int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len) | |
694 | { | |
695 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
696 | int seg; | |
697 | int offset = offset_in_page(gpa); | |
698 | int ret; | |
699 | ||
700 | while ((seg = next_segment(len, offset)) != 0) { | |
701 | ret = kvm_read_guest_page(kvm, gfn, data, offset, seg); | |
702 | if (ret < 0) | |
703 | return ret; | |
704 | offset = 0; | |
705 | len -= seg; | |
706 | data += seg; | |
707 | ++gfn; | |
708 | } | |
709 | return 0; | |
710 | } | |
711 | EXPORT_SYMBOL_GPL(kvm_read_guest); | |
712 | ||
713 | int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data, | |
714 | int offset, int len) | |
715 | { | |
716 | void *page_virt; | |
717 | struct page *page; | |
718 | ||
719 | page = gfn_to_page(kvm, gfn); | |
8a7ae055 IE |
720 | if (is_error_page(page)) { |
721 | kvm_release_page(page); | |
195aefde | 722 | return -EFAULT; |
8a7ae055 | 723 | } |
195aefde IE |
724 | page_virt = kmap_atomic(page, KM_USER0); |
725 | ||
726 | memcpy(page_virt + offset, data, len); | |
727 | ||
728 | kunmap_atomic(page_virt, KM_USER0); | |
729 | mark_page_dirty(kvm, gfn); | |
8a7ae055 | 730 | kvm_release_page(page); |
195aefde IE |
731 | return 0; |
732 | } | |
733 | EXPORT_SYMBOL_GPL(kvm_write_guest_page); | |
734 | ||
735 | int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data, | |
736 | unsigned long len) | |
737 | { | |
738 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
739 | int seg; | |
740 | int offset = offset_in_page(gpa); | |
741 | int ret; | |
742 | ||
743 | while ((seg = next_segment(len, offset)) != 0) { | |
744 | ret = kvm_write_guest_page(kvm, gfn, data, offset, seg); | |
745 | if (ret < 0) | |
746 | return ret; | |
747 | offset = 0; | |
748 | len -= seg; | |
749 | data += seg; | |
750 | ++gfn; | |
751 | } | |
752 | return 0; | |
753 | } | |
754 | ||
755 | int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len) | |
756 | { | |
757 | void *page_virt; | |
758 | struct page *page; | |
759 | ||
760 | page = gfn_to_page(kvm, gfn); | |
8a7ae055 IE |
761 | if (is_error_page(page)) { |
762 | kvm_release_page(page); | |
195aefde | 763 | return -EFAULT; |
8a7ae055 | 764 | } |
195aefde IE |
765 | page_virt = kmap_atomic(page, KM_USER0); |
766 | ||
767 | memset(page_virt + offset, 0, len); | |
768 | ||
769 | kunmap_atomic(page_virt, KM_USER0); | |
8a7ae055 | 770 | kvm_release_page(page); |
195aefde IE |
771 | return 0; |
772 | } | |
773 | EXPORT_SYMBOL_GPL(kvm_clear_guest_page); | |
774 | ||
775 | int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len) | |
776 | { | |
777 | gfn_t gfn = gpa >> PAGE_SHIFT; | |
778 | int seg; | |
779 | int offset = offset_in_page(gpa); | |
780 | int ret; | |
781 | ||
782 | while ((seg = next_segment(len, offset)) != 0) { | |
783 | ret = kvm_clear_guest_page(kvm, gfn, offset, seg); | |
784 | if (ret < 0) | |
785 | return ret; | |
786 | offset = 0; | |
787 | len -= seg; | |
788 | ++gfn; | |
789 | } | |
790 | return 0; | |
791 | } | |
792 | EXPORT_SYMBOL_GPL(kvm_clear_guest); | |
793 | ||
7e9d619d | 794 | /* WARNING: Does not work on aliased pages. */ |
6aa8b732 AK |
795 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
796 | { | |
31389947 | 797 | struct kvm_memory_slot *memslot; |
6aa8b732 | 798 | |
7e9d619d RR |
799 | memslot = __gfn_to_memslot(kvm, gfn); |
800 | if (memslot && memslot->dirty_bitmap) { | |
801 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 802 | |
7e9d619d RR |
803 | /* avoid RMW */ |
804 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
805 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
806 | } |
807 | } | |
808 | ||
e7d5d76c | 809 | int emulator_read_std(unsigned long addr, |
4c690a1e | 810 | void *val, |
6aa8b732 | 811 | unsigned int bytes, |
cebff02b | 812 | struct kvm_vcpu *vcpu) |
6aa8b732 | 813 | { |
6aa8b732 AK |
814 | void *data = val; |
815 | ||
816 | while (bytes) { | |
817 | gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr); | |
818 | unsigned offset = addr & (PAGE_SIZE-1); | |
819 | unsigned tocopy = min(bytes, (unsigned)PAGE_SIZE - offset); | |
195aefde | 820 | int ret; |
6aa8b732 AK |
821 | |
822 | if (gpa == UNMAPPED_GVA) | |
823 | return X86EMUL_PROPAGATE_FAULT; | |
195aefde IE |
824 | ret = kvm_read_guest(vcpu->kvm, gpa, data, tocopy); |
825 | if (ret < 0) | |
6aa8b732 | 826 | return X86EMUL_UNHANDLEABLE; |
6aa8b732 AK |
827 | |
828 | bytes -= tocopy; | |
829 | data += tocopy; | |
830 | addr += tocopy; | |
831 | } | |
832 | ||
833 | return X86EMUL_CONTINUE; | |
834 | } | |
e7d5d76c | 835 | EXPORT_SYMBOL_GPL(emulator_read_std); |
6aa8b732 AK |
836 | |
837 | static int emulator_write_std(unsigned long addr, | |
4c690a1e | 838 | const void *val, |
6aa8b732 | 839 | unsigned int bytes, |
cebff02b | 840 | struct kvm_vcpu *vcpu) |
6aa8b732 | 841 | { |
f0242478 | 842 | pr_unimpl(vcpu, "emulator_write_std: addr %lx n %d\n", addr, bytes); |
6aa8b732 AK |
843 | return X86EMUL_UNHANDLEABLE; |
844 | } | |
845 | ||
97222cc8 ED |
846 | /* |
847 | * Only apic need an MMIO device hook, so shortcut now.. | |
848 | */ | |
849 | static struct kvm_io_device *vcpu_find_pervcpu_dev(struct kvm_vcpu *vcpu, | |
850 | gpa_t addr) | |
851 | { | |
852 | struct kvm_io_device *dev; | |
853 | ||
854 | if (vcpu->apic) { | |
855 | dev = &vcpu->apic->dev; | |
856 | if (dev->in_range(dev, addr)) | |
857 | return dev; | |
858 | } | |
859 | return NULL; | |
860 | } | |
861 | ||
2eeb2e94 GH |
862 | static struct kvm_io_device *vcpu_find_mmio_dev(struct kvm_vcpu *vcpu, |
863 | gpa_t addr) | |
864 | { | |
97222cc8 ED |
865 | struct kvm_io_device *dev; |
866 | ||
867 | dev = vcpu_find_pervcpu_dev(vcpu, addr); | |
868 | if (dev == NULL) | |
869 | dev = kvm_io_bus_find_dev(&vcpu->kvm->mmio_bus, addr); | |
870 | return dev; | |
2eeb2e94 GH |
871 | } |
872 | ||
74906345 ED |
873 | static struct kvm_io_device *vcpu_find_pio_dev(struct kvm_vcpu *vcpu, |
874 | gpa_t addr) | |
875 | { | |
876 | return kvm_io_bus_find_dev(&vcpu->kvm->pio_bus, addr); | |
877 | } | |
878 | ||
6aa8b732 | 879 | static int emulator_read_emulated(unsigned long addr, |
4c690a1e | 880 | void *val, |
6aa8b732 | 881 | unsigned int bytes, |
cebff02b | 882 | struct kvm_vcpu *vcpu) |
6aa8b732 | 883 | { |
2eeb2e94 GH |
884 | struct kvm_io_device *mmio_dev; |
885 | gpa_t gpa; | |
6aa8b732 AK |
886 | |
887 | if (vcpu->mmio_read_completed) { | |
888 | memcpy(val, vcpu->mmio_data, bytes); | |
889 | vcpu->mmio_read_completed = 0; | |
890 | return X86EMUL_CONTINUE; | |
cebff02b | 891 | } else if (emulator_read_std(addr, val, bytes, vcpu) |
6aa8b732 AK |
892 | == X86EMUL_CONTINUE) |
893 | return X86EMUL_CONTINUE; | |
d27d4aca | 894 | |
2eeb2e94 GH |
895 | gpa = vcpu->mmu.gva_to_gpa(vcpu, addr); |
896 | if (gpa == UNMAPPED_GVA) | |
897 | return X86EMUL_PROPAGATE_FAULT; | |
6aa8b732 | 898 | |
2eeb2e94 GH |
899 | /* |
900 | * Is this MMIO handled locally? | |
901 | */ | |
902 | mmio_dev = vcpu_find_mmio_dev(vcpu, gpa); | |
903 | if (mmio_dev) { | |
904 | kvm_iodevice_read(mmio_dev, gpa, bytes, val); | |
905 | return X86EMUL_CONTINUE; | |
6aa8b732 | 906 | } |
2eeb2e94 GH |
907 | |
908 | vcpu->mmio_needed = 1; | |
909 | vcpu->mmio_phys_addr = gpa; | |
910 | vcpu->mmio_size = bytes; | |
911 | vcpu->mmio_is_write = 0; | |
912 | ||
913 | return X86EMUL_UNHANDLEABLE; | |
6aa8b732 AK |
914 | } |
915 | ||
da4a00f0 | 916 | static int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa, |
4c690a1e | 917 | const void *val, int bytes) |
da4a00f0 | 918 | { |
195aefde | 919 | int ret; |
da4a00f0 | 920 | |
195aefde IE |
921 | ret = kvm_write_guest(vcpu->kvm, gpa, val, bytes); |
922 | if (ret < 0) | |
da4a00f0 | 923 | return 0; |
fe551881 | 924 | kvm_mmu_pte_write(vcpu, gpa, val, bytes); |
da4a00f0 AK |
925 | return 1; |
926 | } | |
927 | ||
b0fcd903 AK |
928 | static int emulator_write_emulated_onepage(unsigned long addr, |
929 | const void *val, | |
930 | unsigned int bytes, | |
cebff02b | 931 | struct kvm_vcpu *vcpu) |
6aa8b732 | 932 | { |
2eeb2e94 GH |
933 | struct kvm_io_device *mmio_dev; |
934 | gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr); | |
6aa8b732 | 935 | |
c9047f53 | 936 | if (gpa == UNMAPPED_GVA) { |
cbdd1bea | 937 | kvm_x86_ops->inject_page_fault(vcpu, addr, 2); |
6aa8b732 | 938 | return X86EMUL_PROPAGATE_FAULT; |
c9047f53 | 939 | } |
6aa8b732 | 940 | |
da4a00f0 AK |
941 | if (emulator_write_phys(vcpu, gpa, val, bytes)) |
942 | return X86EMUL_CONTINUE; | |
943 | ||
2eeb2e94 GH |
944 | /* |
945 | * Is this MMIO handled locally? | |
946 | */ | |
947 | mmio_dev = vcpu_find_mmio_dev(vcpu, gpa); | |
948 | if (mmio_dev) { | |
949 | kvm_iodevice_write(mmio_dev, gpa, bytes, val); | |
950 | return X86EMUL_CONTINUE; | |
951 | } | |
952 | ||
6aa8b732 AK |
953 | vcpu->mmio_needed = 1; |
954 | vcpu->mmio_phys_addr = gpa; | |
955 | vcpu->mmio_size = bytes; | |
956 | vcpu->mmio_is_write = 1; | |
4c690a1e | 957 | memcpy(vcpu->mmio_data, val, bytes); |
6aa8b732 AK |
958 | |
959 | return X86EMUL_CONTINUE; | |
960 | } | |
961 | ||
e7d5d76c | 962 | int emulator_write_emulated(unsigned long addr, |
b0fcd903 AK |
963 | const void *val, |
964 | unsigned int bytes, | |
cebff02b | 965 | struct kvm_vcpu *vcpu) |
b0fcd903 AK |
966 | { |
967 | /* Crossing a page boundary? */ | |
968 | if (((addr + bytes - 1) ^ addr) & PAGE_MASK) { | |
969 | int rc, now; | |
970 | ||
971 | now = -addr & ~PAGE_MASK; | |
cebff02b | 972 | rc = emulator_write_emulated_onepage(addr, val, now, vcpu); |
b0fcd903 AK |
973 | if (rc != X86EMUL_CONTINUE) |
974 | return rc; | |
975 | addr += now; | |
976 | val += now; | |
977 | bytes -= now; | |
978 | } | |
cebff02b | 979 | return emulator_write_emulated_onepage(addr, val, bytes, vcpu); |
b0fcd903 | 980 | } |
e7d5d76c | 981 | EXPORT_SYMBOL_GPL(emulator_write_emulated); |
b0fcd903 | 982 | |
6aa8b732 | 983 | static int emulator_cmpxchg_emulated(unsigned long addr, |
4c690a1e AK |
984 | const void *old, |
985 | const void *new, | |
6aa8b732 | 986 | unsigned int bytes, |
cebff02b | 987 | struct kvm_vcpu *vcpu) |
6aa8b732 AK |
988 | { |
989 | static int reported; | |
990 | ||
991 | if (!reported) { | |
992 | reported = 1; | |
993 | printk(KERN_WARNING "kvm: emulating exchange as write\n"); | |
994 | } | |
cebff02b | 995 | return emulator_write_emulated(addr, new, bytes, vcpu); |
6aa8b732 AK |
996 | } |
997 | ||
998 | static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg) | |
999 | { | |
cbdd1bea | 1000 | return kvm_x86_ops->get_segment_base(vcpu, seg); |
6aa8b732 AK |
1001 | } |
1002 | ||
1003 | int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address) | |
1004 | { | |
6aa8b732 AK |
1005 | return X86EMUL_CONTINUE; |
1006 | } | |
1007 | ||
1008 | int emulate_clts(struct kvm_vcpu *vcpu) | |
1009 | { | |
404fb881 | 1010 | kvm_x86_ops->set_cr0(vcpu, vcpu->cr0 & ~X86_CR0_TS); |
6aa8b732 AK |
1011 | return X86EMUL_CONTINUE; |
1012 | } | |
1013 | ||
d77c26fc | 1014 | int emulator_get_dr(struct x86_emulate_ctxt *ctxt, int dr, unsigned long *dest) |
6aa8b732 AK |
1015 | { |
1016 | struct kvm_vcpu *vcpu = ctxt->vcpu; | |
1017 | ||
1018 | switch (dr) { | |
1019 | case 0 ... 3: | |
cbdd1bea | 1020 | *dest = kvm_x86_ops->get_dr(vcpu, dr); |
6aa8b732 AK |
1021 | return X86EMUL_CONTINUE; |
1022 | default: | |
f0242478 | 1023 | pr_unimpl(vcpu, "%s: unexpected dr %u\n", __FUNCTION__, dr); |
6aa8b732 AK |
1024 | return X86EMUL_UNHANDLEABLE; |
1025 | } | |
1026 | } | |
1027 | ||
1028 | int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr, unsigned long value) | |
1029 | { | |
1030 | unsigned long mask = (ctxt->mode == X86EMUL_MODE_PROT64) ? ~0ULL : ~0U; | |
1031 | int exception; | |
1032 | ||
cbdd1bea | 1033 | kvm_x86_ops->set_dr(ctxt->vcpu, dr, value & mask, &exception); |
6aa8b732 AK |
1034 | if (exception) { |
1035 | /* FIXME: better handling */ | |
1036 | return X86EMUL_UNHANDLEABLE; | |
1037 | } | |
1038 | return X86EMUL_CONTINUE; | |
1039 | } | |
1040 | ||
054b1369 | 1041 | void kvm_report_emulation_failure(struct kvm_vcpu *vcpu, const char *context) |
6aa8b732 AK |
1042 | { |
1043 | static int reported; | |
1044 | u8 opcodes[4]; | |
054b1369 | 1045 | unsigned long rip = vcpu->rip; |
6aa8b732 AK |
1046 | unsigned long rip_linear; |
1047 | ||
054b1369 | 1048 | rip_linear = rip + get_segment_base(vcpu, VCPU_SREG_CS); |
6aa8b732 AK |
1049 | |
1050 | if (reported) | |
1051 | return; | |
1052 | ||
054b1369 | 1053 | emulator_read_std(rip_linear, (void *)opcodes, 4, vcpu); |
6aa8b732 | 1054 | |
054b1369 AK |
1055 | printk(KERN_ERR "emulation failed (%s) rip %lx %02x %02x %02x %02x\n", |
1056 | context, rip, opcodes[0], opcodes[1], opcodes[2], opcodes[3]); | |
6aa8b732 AK |
1057 | reported = 1; |
1058 | } | |
054b1369 | 1059 | EXPORT_SYMBOL_GPL(kvm_report_emulation_failure); |
6aa8b732 AK |
1060 | |
1061 | struct x86_emulate_ops emulate_ops = { | |
1062 | .read_std = emulator_read_std, | |
1063 | .write_std = emulator_write_std, | |
1064 | .read_emulated = emulator_read_emulated, | |
1065 | .write_emulated = emulator_write_emulated, | |
1066 | .cmpxchg_emulated = emulator_cmpxchg_emulated, | |
1067 | }; | |
1068 | ||
1069 | int emulate_instruction(struct kvm_vcpu *vcpu, | |
1070 | struct kvm_run *run, | |
1071 | unsigned long cr2, | |
3427318f LV |
1072 | u16 error_code, |
1073 | int no_decode) | |
6aa8b732 | 1074 | { |
a22436b7 | 1075 | int r; |
6aa8b732 | 1076 | |
e7df56e4 | 1077 | vcpu->mmio_fault_cr2 = cr2; |
cbdd1bea | 1078 | kvm_x86_ops->cache_regs(vcpu); |
6aa8b732 | 1079 | |
6aa8b732 | 1080 | vcpu->mmio_is_write = 0; |
e70669ab | 1081 | vcpu->pio.string = 0; |
3427318f LV |
1082 | |
1083 | if (!no_decode) { | |
1084 | int cs_db, cs_l; | |
1085 | kvm_x86_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l); | |
1086 | ||
1087 | vcpu->emulate_ctxt.vcpu = vcpu; | |
1088 | vcpu->emulate_ctxt.eflags = kvm_x86_ops->get_rflags(vcpu); | |
1089 | vcpu->emulate_ctxt.cr2 = cr2; | |
1090 | vcpu->emulate_ctxt.mode = | |
1091 | (vcpu->emulate_ctxt.eflags & X86_EFLAGS_VM) | |
1092 | ? X86EMUL_MODE_REAL : cs_l | |
1093 | ? X86EMUL_MODE_PROT64 : cs_db | |
1094 | ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16; | |
1095 | ||
1096 | if (vcpu->emulate_ctxt.mode == X86EMUL_MODE_PROT64) { | |
1097 | vcpu->emulate_ctxt.cs_base = 0; | |
1098 | vcpu->emulate_ctxt.ds_base = 0; | |
1099 | vcpu->emulate_ctxt.es_base = 0; | |
1100 | vcpu->emulate_ctxt.ss_base = 0; | |
1101 | } else { | |
1102 | vcpu->emulate_ctxt.cs_base = | |
1103 | get_segment_base(vcpu, VCPU_SREG_CS); | |
1104 | vcpu->emulate_ctxt.ds_base = | |
1105 | get_segment_base(vcpu, VCPU_SREG_DS); | |
1106 | vcpu->emulate_ctxt.es_base = | |
1107 | get_segment_base(vcpu, VCPU_SREG_ES); | |
1108 | vcpu->emulate_ctxt.ss_base = | |
1109 | get_segment_base(vcpu, VCPU_SREG_SS); | |
1110 | } | |
1111 | ||
1112 | vcpu->emulate_ctxt.gs_base = | |
1113 | get_segment_base(vcpu, VCPU_SREG_GS); | |
1114 | vcpu->emulate_ctxt.fs_base = | |
1115 | get_segment_base(vcpu, VCPU_SREG_FS); | |
1116 | ||
1117 | r = x86_decode_insn(&vcpu->emulate_ctxt, &emulate_ops); | |
a22436b7 LV |
1118 | if (r) { |
1119 | if (kvm_mmu_unprotect_page_virt(vcpu, cr2)) | |
1120 | return EMULATE_DONE; | |
1121 | return EMULATE_FAIL; | |
1122 | } | |
3427318f LV |
1123 | } |
1124 | ||
a22436b7 | 1125 | r = x86_emulate_insn(&vcpu->emulate_ctxt, &emulate_ops); |
1be3aa47 | 1126 | |
e70669ab LV |
1127 | if (vcpu->pio.string) |
1128 | return EMULATE_DO_MMIO; | |
6aa8b732 AK |
1129 | |
1130 | if ((r || vcpu->mmio_is_write) && run) { | |
8fc0d085 | 1131 | run->exit_reason = KVM_EXIT_MMIO; |
6aa8b732 AK |
1132 | run->mmio.phys_addr = vcpu->mmio_phys_addr; |
1133 | memcpy(run->mmio.data, vcpu->mmio_data, 8); | |
1134 | run->mmio.len = vcpu->mmio_size; | |
1135 | run->mmio.is_write = vcpu->mmio_is_write; | |
1136 | } | |
1137 | ||
1138 | if (r) { | |
a436036b AK |
1139 | if (kvm_mmu_unprotect_page_virt(vcpu, cr2)) |
1140 | return EMULATE_DONE; | |
6aa8b732 | 1141 | if (!vcpu->mmio_needed) { |
054b1369 | 1142 | kvm_report_emulation_failure(vcpu, "mmio"); |
6aa8b732 AK |
1143 | return EMULATE_FAIL; |
1144 | } | |
1145 | return EMULATE_DO_MMIO; | |
1146 | } | |
1147 | ||
cbdd1bea | 1148 | kvm_x86_ops->decache_regs(vcpu); |
3427318f | 1149 | kvm_x86_ops->set_rflags(vcpu, vcpu->emulate_ctxt.eflags); |
6aa8b732 | 1150 | |
02c83209 AK |
1151 | if (vcpu->mmio_is_write) { |
1152 | vcpu->mmio_needed = 0; | |
6aa8b732 | 1153 | return EMULATE_DO_MMIO; |
02c83209 | 1154 | } |
6aa8b732 AK |
1155 | |
1156 | return EMULATE_DONE; | |
1157 | } | |
1158 | EXPORT_SYMBOL_GPL(emulate_instruction); | |
1159 | ||
b6958ce4 ED |
1160 | /* |
1161 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1162 | */ | |
c5ec1534 | 1163 | static void kvm_vcpu_block(struct kvm_vcpu *vcpu) |
d3bef15f | 1164 | { |
b6958ce4 ED |
1165 | DECLARE_WAITQUEUE(wait, current); |
1166 | ||
1167 | add_wait_queue(&vcpu->wq, &wait); | |
1168 | ||
1169 | /* | |
1170 | * We will block until either an interrupt or a signal wakes us up | |
1171 | */ | |
c5ec1534 HQ |
1172 | while (!kvm_cpu_has_interrupt(vcpu) |
1173 | && !signal_pending(current) | |
1174 | && vcpu->mp_state != VCPU_MP_STATE_RUNNABLE | |
1175 | && vcpu->mp_state != VCPU_MP_STATE_SIPI_RECEIVED) { | |
b6958ce4 ED |
1176 | set_current_state(TASK_INTERRUPTIBLE); |
1177 | vcpu_put(vcpu); | |
1178 | schedule(); | |
1179 | vcpu_load(vcpu); | |
1180 | } | |
d3bef15f | 1181 | |
c5ec1534 | 1182 | __set_current_state(TASK_RUNNING); |
b6958ce4 | 1183 | remove_wait_queue(&vcpu->wq, &wait); |
b6958ce4 ED |
1184 | } |
1185 | ||
1186 | int kvm_emulate_halt(struct kvm_vcpu *vcpu) | |
1187 | { | |
d3bef15f | 1188 | ++vcpu->stat.halt_exits; |
b6958ce4 | 1189 | if (irqchip_in_kernel(vcpu->kvm)) { |
c5ec1534 HQ |
1190 | vcpu->mp_state = VCPU_MP_STATE_HALTED; |
1191 | kvm_vcpu_block(vcpu); | |
1192 | if (vcpu->mp_state != VCPU_MP_STATE_RUNNABLE) | |
1193 | return -EINTR; | |
b6958ce4 ED |
1194 | return 1; |
1195 | } else { | |
1196 | vcpu->run->exit_reason = KVM_EXIT_HLT; | |
1197 | return 0; | |
1198 | } | |
d3bef15f AK |
1199 | } |
1200 | EXPORT_SYMBOL_GPL(kvm_emulate_halt); | |
1201 | ||
7aa81cc0 | 1202 | int kvm_emulate_hypercall(struct kvm_vcpu *vcpu) |
270fd9b9 | 1203 | { |
7aa81cc0 | 1204 | unsigned long nr, a0, a1, a2, a3, ret; |
270fd9b9 | 1205 | |
cbdd1bea | 1206 | kvm_x86_ops->cache_regs(vcpu); |
7aa81cc0 AL |
1207 | |
1208 | nr = vcpu->regs[VCPU_REGS_RAX]; | |
1209 | a0 = vcpu->regs[VCPU_REGS_RBX]; | |
1210 | a1 = vcpu->regs[VCPU_REGS_RCX]; | |
1211 | a2 = vcpu->regs[VCPU_REGS_RDX]; | |
1212 | a3 = vcpu->regs[VCPU_REGS_RSI]; | |
1213 | ||
1214 | if (!is_long_mode(vcpu)) { | |
1215 | nr &= 0xFFFFFFFF; | |
1216 | a0 &= 0xFFFFFFFF; | |
1217 | a1 &= 0xFFFFFFFF; | |
1218 | a2 &= 0xFFFFFFFF; | |
1219 | a3 &= 0xFFFFFFFF; | |
270fd9b9 | 1220 | } |
7aa81cc0 | 1221 | |
270fd9b9 AK |
1222 | switch (nr) { |
1223 | default: | |
7aa81cc0 AL |
1224 | ret = -KVM_ENOSYS; |
1225 | break; | |
270fd9b9 AK |
1226 | } |
1227 | vcpu->regs[VCPU_REGS_RAX] = ret; | |
cbdd1bea | 1228 | kvm_x86_ops->decache_regs(vcpu); |
7aa81cc0 AL |
1229 | return 0; |
1230 | } | |
1231 | EXPORT_SYMBOL_GPL(kvm_emulate_hypercall); | |
1232 | ||
1233 | int kvm_fix_hypercall(struct kvm_vcpu *vcpu) | |
1234 | { | |
1235 | char instruction[3]; | |
1236 | int ret = 0; | |
1237 | ||
1238 | mutex_lock(&vcpu->kvm->lock); | |
1239 | ||
1240 | /* | |
1241 | * Blow out the MMU to ensure that no other VCPU has an active mapping | |
1242 | * to ensure that the updated hypercall appears atomically across all | |
1243 | * VCPUs. | |
1244 | */ | |
1245 | kvm_mmu_zap_all(vcpu->kvm); | |
1246 | ||
1247 | kvm_x86_ops->cache_regs(vcpu); | |
1248 | kvm_x86_ops->patch_hypercall(vcpu, instruction); | |
1249 | if (emulator_write_emulated(vcpu->rip, instruction, 3, vcpu) | |
1250 | != X86EMUL_CONTINUE) | |
1251 | ret = -EFAULT; | |
1252 | ||
1253 | mutex_unlock(&vcpu->kvm->lock); | |
1254 | ||
1255 | return ret; | |
270fd9b9 | 1256 | } |
270fd9b9 | 1257 | |
6aa8b732 AK |
1258 | static u64 mk_cr_64(u64 curr_cr, u32 new_val) |
1259 | { | |
1260 | return (curr_cr & ~((1ULL << 32) - 1)) | new_val; | |
1261 | } | |
1262 | ||
1263 | void realmode_lgdt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base) | |
1264 | { | |
1265 | struct descriptor_table dt = { limit, base }; | |
1266 | ||
cbdd1bea | 1267 | kvm_x86_ops->set_gdt(vcpu, &dt); |
6aa8b732 AK |
1268 | } |
1269 | ||
1270 | void realmode_lidt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base) | |
1271 | { | |
1272 | struct descriptor_table dt = { limit, base }; | |
1273 | ||
cbdd1bea | 1274 | kvm_x86_ops->set_idt(vcpu, &dt); |
6aa8b732 AK |
1275 | } |
1276 | ||
1277 | void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw, | |
1278 | unsigned long *rflags) | |
1279 | { | |
1280 | lmsw(vcpu, msw); | |
cbdd1bea | 1281 | *rflags = kvm_x86_ops->get_rflags(vcpu); |
6aa8b732 AK |
1282 | } |
1283 | ||
1284 | unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr) | |
1285 | { | |
cbdd1bea | 1286 | kvm_x86_ops->decache_cr4_guest_bits(vcpu); |
6aa8b732 AK |
1287 | switch (cr) { |
1288 | case 0: | |
1289 | return vcpu->cr0; | |
1290 | case 2: | |
1291 | return vcpu->cr2; | |
1292 | case 3: | |
1293 | return vcpu->cr3; | |
1294 | case 4: | |
1295 | return vcpu->cr4; | |
1296 | default: | |
1297 | vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr); | |
1298 | return 0; | |
1299 | } | |
1300 | } | |
1301 | ||
1302 | void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long val, | |
1303 | unsigned long *rflags) | |
1304 | { | |
1305 | switch (cr) { | |
1306 | case 0: | |
1307 | set_cr0(vcpu, mk_cr_64(vcpu->cr0, val)); | |
cbdd1bea | 1308 | *rflags = kvm_x86_ops->get_rflags(vcpu); |
6aa8b732 AK |
1309 | break; |
1310 | case 2: | |
1311 | vcpu->cr2 = val; | |
1312 | break; | |
1313 | case 3: | |
1314 | set_cr3(vcpu, val); | |
1315 | break; | |
1316 | case 4: | |
1317 | set_cr4(vcpu, mk_cr_64(vcpu->cr4, val)); | |
1318 | break; | |
1319 | default: | |
1320 | vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr); | |
1321 | } | |
1322 | } | |
1323 | ||
3bab1f5d AK |
1324 | int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) |
1325 | { | |
1326 | u64 data; | |
1327 | ||
1328 | switch (msr) { | |
1329 | case 0xc0010010: /* SYSCFG */ | |
1330 | case 0xc0010015: /* HWCR */ | |
1331 | case MSR_IA32_PLATFORM_ID: | |
1332 | case MSR_IA32_P5_MC_ADDR: | |
1333 | case MSR_IA32_P5_MC_TYPE: | |
1334 | case MSR_IA32_MC0_CTL: | |
1335 | case MSR_IA32_MCG_STATUS: | |
1336 | case MSR_IA32_MCG_CAP: | |
1337 | case MSR_IA32_MC0_MISC: | |
1338 | case MSR_IA32_MC0_MISC+4: | |
1339 | case MSR_IA32_MC0_MISC+8: | |
1340 | case MSR_IA32_MC0_MISC+12: | |
1341 | case MSR_IA32_MC0_MISC+16: | |
1342 | case MSR_IA32_UCODE_REV: | |
a8d13ea2 | 1343 | case MSR_IA32_PERF_STATUS: |
2dc7094b | 1344 | case MSR_IA32_EBL_CR_POWERON: |
3bab1f5d AK |
1345 | /* MTRR registers */ |
1346 | case 0xfe: | |
1347 | case 0x200 ... 0x2ff: | |
1348 | data = 0; | |
1349 | break; | |
a8d13ea2 AK |
1350 | case 0xcd: /* fsb frequency */ |
1351 | data = 3; | |
1352 | break; | |
3bab1f5d | 1353 | case MSR_IA32_APICBASE: |
7017fc3d | 1354 | data = kvm_get_apic_base(vcpu); |
3bab1f5d | 1355 | break; |
6f00e68f AK |
1356 | case MSR_IA32_MISC_ENABLE: |
1357 | data = vcpu->ia32_misc_enable_msr; | |
1358 | break; | |
3bab1f5d AK |
1359 | #ifdef CONFIG_X86_64 |
1360 | case MSR_EFER: | |
1361 | data = vcpu->shadow_efer; | |
1362 | break; | |
1363 | #endif | |
1364 | default: | |
f0242478 | 1365 | pr_unimpl(vcpu, "unhandled rdmsr: 0x%x\n", msr); |
3bab1f5d AK |
1366 | return 1; |
1367 | } | |
1368 | *pdata = data; | |
1369 | return 0; | |
1370 | } | |
1371 | EXPORT_SYMBOL_GPL(kvm_get_msr_common); | |
1372 | ||
6aa8b732 AK |
1373 | /* |
1374 | * Reads an msr value (of 'msr_index') into 'pdata'. | |
1375 | * Returns 0 on success, non-0 otherwise. | |
1376 | * Assumes vcpu_load() was already called. | |
1377 | */ | |
35f3f286 | 1378 | int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata) |
6aa8b732 | 1379 | { |
cbdd1bea | 1380 | return kvm_x86_ops->get_msr(vcpu, msr_index, pdata); |
6aa8b732 AK |
1381 | } |
1382 | ||
05b3e0c2 | 1383 | #ifdef CONFIG_X86_64 |
6aa8b732 | 1384 | |
3bab1f5d | 1385 | static void set_efer(struct kvm_vcpu *vcpu, u64 efer) |
6aa8b732 | 1386 | { |
6aa8b732 AK |
1387 | if (efer & EFER_RESERVED_BITS) { |
1388 | printk(KERN_DEBUG "set_efer: 0x%llx #GP, reserved bits\n", | |
1389 | efer); | |
1390 | inject_gp(vcpu); | |
1391 | return; | |
1392 | } | |
1393 | ||
1394 | if (is_paging(vcpu) | |
1395 | && (vcpu->shadow_efer & EFER_LME) != (efer & EFER_LME)) { | |
1396 | printk(KERN_DEBUG "set_efer: #GP, change LME while paging\n"); | |
1397 | inject_gp(vcpu); | |
1398 | return; | |
1399 | } | |
1400 | ||
cbdd1bea | 1401 | kvm_x86_ops->set_efer(vcpu, efer); |
7725f0ba | 1402 | |
6aa8b732 AK |
1403 | efer &= ~EFER_LMA; |
1404 | efer |= vcpu->shadow_efer & EFER_LMA; | |
1405 | ||
1406 | vcpu->shadow_efer = efer; | |
6aa8b732 | 1407 | } |
6aa8b732 AK |
1408 | |
1409 | #endif | |
1410 | ||
3bab1f5d AK |
1411 | int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data) |
1412 | { | |
1413 | switch (msr) { | |
1414 | #ifdef CONFIG_X86_64 | |
1415 | case MSR_EFER: | |
1416 | set_efer(vcpu, data); | |
1417 | break; | |
1418 | #endif | |
1419 | case MSR_IA32_MC0_STATUS: | |
f0242478 | 1420 | pr_unimpl(vcpu, "%s: MSR_IA32_MC0_STATUS 0x%llx, nop\n", |
3bab1f5d AK |
1421 | __FUNCTION__, data); |
1422 | break; | |
0e5bf0d0 | 1423 | case MSR_IA32_MCG_STATUS: |
f0242478 | 1424 | pr_unimpl(vcpu, "%s: MSR_IA32_MCG_STATUS 0x%llx, nop\n", |
0e5bf0d0 SK |
1425 | __FUNCTION__, data); |
1426 | break; | |
3bab1f5d AK |
1427 | case MSR_IA32_UCODE_REV: |
1428 | case MSR_IA32_UCODE_WRITE: | |
1429 | case 0x200 ... 0x2ff: /* MTRRs */ | |
1430 | break; | |
1431 | case MSR_IA32_APICBASE: | |
7017fc3d | 1432 | kvm_set_apic_base(vcpu, data); |
3bab1f5d | 1433 | break; |
6f00e68f AK |
1434 | case MSR_IA32_MISC_ENABLE: |
1435 | vcpu->ia32_misc_enable_msr = data; | |
1436 | break; | |
3bab1f5d | 1437 | default: |
f0242478 | 1438 | pr_unimpl(vcpu, "unhandled wrmsr: 0x%x\n", msr); |
3bab1f5d AK |
1439 | return 1; |
1440 | } | |
1441 | return 0; | |
1442 | } | |
1443 | EXPORT_SYMBOL_GPL(kvm_set_msr_common); | |
1444 | ||
6aa8b732 AK |
1445 | /* |
1446 | * Writes msr value into into the appropriate "register". | |
1447 | * Returns 0 on success, non-0 otherwise. | |
1448 | * Assumes vcpu_load() was already called. | |
1449 | */ | |
35f3f286 | 1450 | int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data) |
6aa8b732 | 1451 | { |
cbdd1bea | 1452 | return kvm_x86_ops->set_msr(vcpu, msr_index, data); |
6aa8b732 AK |
1453 | } |
1454 | ||
1455 | void kvm_resched(struct kvm_vcpu *vcpu) | |
1456 | { | |
3fca0365 YD |
1457 | if (!need_resched()) |
1458 | return; | |
6aa8b732 | 1459 | cond_resched(); |
6aa8b732 AK |
1460 | } |
1461 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1462 | ||
06465c5a AK |
1463 | void kvm_emulate_cpuid(struct kvm_vcpu *vcpu) |
1464 | { | |
1465 | int i; | |
1466 | u32 function; | |
1467 | struct kvm_cpuid_entry *e, *best; | |
1468 | ||
cbdd1bea | 1469 | kvm_x86_ops->cache_regs(vcpu); |
06465c5a AK |
1470 | function = vcpu->regs[VCPU_REGS_RAX]; |
1471 | vcpu->regs[VCPU_REGS_RAX] = 0; | |
1472 | vcpu->regs[VCPU_REGS_RBX] = 0; | |
1473 | vcpu->regs[VCPU_REGS_RCX] = 0; | |
1474 | vcpu->regs[VCPU_REGS_RDX] = 0; | |
1475 | best = NULL; | |
1476 | for (i = 0; i < vcpu->cpuid_nent; ++i) { | |
1477 | e = &vcpu->cpuid_entries[i]; | |
1478 | if (e->function == function) { | |
1479 | best = e; | |
1480 | break; | |
1481 | } | |
1482 | /* | |
1483 | * Both basic or both extended? | |
1484 | */ | |
1485 | if (((e->function ^ function) & 0x80000000) == 0) | |
1486 | if (!best || e->function > best->function) | |
1487 | best = e; | |
1488 | } | |
1489 | if (best) { | |
1490 | vcpu->regs[VCPU_REGS_RAX] = best->eax; | |
1491 | vcpu->regs[VCPU_REGS_RBX] = best->ebx; | |
1492 | vcpu->regs[VCPU_REGS_RCX] = best->ecx; | |
1493 | vcpu->regs[VCPU_REGS_RDX] = best->edx; | |
1494 | } | |
cbdd1bea CE |
1495 | kvm_x86_ops->decache_regs(vcpu); |
1496 | kvm_x86_ops->skip_emulated_instruction(vcpu); | |
06465c5a AK |
1497 | } |
1498 | EXPORT_SYMBOL_GPL(kvm_emulate_cpuid); | |
1499 | ||
039576c0 | 1500 | static int pio_copy_data(struct kvm_vcpu *vcpu) |
46fc1477 | 1501 | { |
039576c0 AK |
1502 | void *p = vcpu->pio_data; |
1503 | void *q; | |
1504 | unsigned bytes; | |
1505 | int nr_pages = vcpu->pio.guest_pages[1] ? 2 : 1; | |
1506 | ||
039576c0 AK |
1507 | q = vmap(vcpu->pio.guest_pages, nr_pages, VM_READ|VM_WRITE, |
1508 | PAGE_KERNEL); | |
1509 | if (!q) { | |
039576c0 AK |
1510 | free_pio_guest_pages(vcpu); |
1511 | return -ENOMEM; | |
1512 | } | |
1513 | q += vcpu->pio.guest_page_offset; | |
1514 | bytes = vcpu->pio.size * vcpu->pio.cur_count; | |
1515 | if (vcpu->pio.in) | |
1516 | memcpy(q, p, bytes); | |
1517 | else | |
1518 | memcpy(p, q, bytes); | |
1519 | q -= vcpu->pio.guest_page_offset; | |
1520 | vunmap(q); | |
039576c0 AK |
1521 | free_pio_guest_pages(vcpu); |
1522 | return 0; | |
1523 | } | |
1524 | ||
1525 | static int complete_pio(struct kvm_vcpu *vcpu) | |
1526 | { | |
1527 | struct kvm_pio_request *io = &vcpu->pio; | |
46fc1477 | 1528 | long delta; |
039576c0 | 1529 | int r; |
46fc1477 | 1530 | |
cbdd1bea | 1531 | kvm_x86_ops->cache_regs(vcpu); |
46fc1477 AK |
1532 | |
1533 | if (!io->string) { | |
039576c0 AK |
1534 | if (io->in) |
1535 | memcpy(&vcpu->regs[VCPU_REGS_RAX], vcpu->pio_data, | |
46fc1477 AK |
1536 | io->size); |
1537 | } else { | |
039576c0 AK |
1538 | if (io->in) { |
1539 | r = pio_copy_data(vcpu); | |
1540 | if (r) { | |
cbdd1bea | 1541 | kvm_x86_ops->cache_regs(vcpu); |
039576c0 AK |
1542 | return r; |
1543 | } | |
1544 | } | |
1545 | ||
46fc1477 AK |
1546 | delta = 1; |
1547 | if (io->rep) { | |
039576c0 | 1548 | delta *= io->cur_count; |
46fc1477 AK |
1549 | /* |
1550 | * The size of the register should really depend on | |
1551 | * current address size. | |
1552 | */ | |
1553 | vcpu->regs[VCPU_REGS_RCX] -= delta; | |
1554 | } | |
039576c0 | 1555 | if (io->down) |
46fc1477 AK |
1556 | delta = -delta; |
1557 | delta *= io->size; | |
039576c0 | 1558 | if (io->in) |
46fc1477 AK |
1559 | vcpu->regs[VCPU_REGS_RDI] += delta; |
1560 | else | |
1561 | vcpu->regs[VCPU_REGS_RSI] += delta; | |
1562 | } | |
1563 | ||
cbdd1bea | 1564 | kvm_x86_ops->decache_regs(vcpu); |
46fc1477 | 1565 | |
039576c0 AK |
1566 | io->count -= io->cur_count; |
1567 | io->cur_count = 0; | |
1568 | ||
039576c0 | 1569 | return 0; |
46fc1477 AK |
1570 | } |
1571 | ||
65619eb5 ED |
1572 | static void kernel_pio(struct kvm_io_device *pio_dev, |
1573 | struct kvm_vcpu *vcpu, | |
1574 | void *pd) | |
74906345 ED |
1575 | { |
1576 | /* TODO: String I/O for in kernel device */ | |
1577 | ||
9cf98828 | 1578 | mutex_lock(&vcpu->kvm->lock); |
74906345 ED |
1579 | if (vcpu->pio.in) |
1580 | kvm_iodevice_read(pio_dev, vcpu->pio.port, | |
1581 | vcpu->pio.size, | |
65619eb5 | 1582 | pd); |
74906345 ED |
1583 | else |
1584 | kvm_iodevice_write(pio_dev, vcpu->pio.port, | |
1585 | vcpu->pio.size, | |
65619eb5 | 1586 | pd); |
9cf98828 | 1587 | mutex_unlock(&vcpu->kvm->lock); |
65619eb5 ED |
1588 | } |
1589 | ||
1590 | static void pio_string_write(struct kvm_io_device *pio_dev, | |
1591 | struct kvm_vcpu *vcpu) | |
1592 | { | |
1593 | struct kvm_pio_request *io = &vcpu->pio; | |
1594 | void *pd = vcpu->pio_data; | |
1595 | int i; | |
1596 | ||
9cf98828 | 1597 | mutex_lock(&vcpu->kvm->lock); |
65619eb5 ED |
1598 | for (i = 0; i < io->cur_count; i++) { |
1599 | kvm_iodevice_write(pio_dev, io->port, | |
1600 | io->size, | |
1601 | pd); | |
1602 | pd += io->size; | |
1603 | } | |
9cf98828 | 1604 | mutex_unlock(&vcpu->kvm->lock); |
74906345 ED |
1605 | } |
1606 | ||
d77c26fc | 1607 | int kvm_emulate_pio(struct kvm_vcpu *vcpu, struct kvm_run *run, int in, |
3090dd73 LV |
1608 | int size, unsigned port) |
1609 | { | |
1610 | struct kvm_io_device *pio_dev; | |
1611 | ||
1612 | vcpu->run->exit_reason = KVM_EXIT_IO; | |
1613 | vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; | |
1614 | vcpu->run->io.size = vcpu->pio.size = size; | |
1615 | vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; | |
1616 | vcpu->run->io.count = vcpu->pio.count = vcpu->pio.cur_count = 1; | |
1617 | vcpu->run->io.port = vcpu->pio.port = port; | |
1618 | vcpu->pio.in = in; | |
1619 | vcpu->pio.string = 0; | |
1620 | vcpu->pio.down = 0; | |
1621 | vcpu->pio.guest_page_offset = 0; | |
1622 | vcpu->pio.rep = 0; | |
1623 | ||
cbdd1bea | 1624 | kvm_x86_ops->cache_regs(vcpu); |
3090dd73 | 1625 | memcpy(vcpu->pio_data, &vcpu->regs[VCPU_REGS_RAX], 4); |
cbdd1bea | 1626 | kvm_x86_ops->decache_regs(vcpu); |
3090dd73 | 1627 | |
0967b7bf AK |
1628 | kvm_x86_ops->skip_emulated_instruction(vcpu); |
1629 | ||
3090dd73 LV |
1630 | pio_dev = vcpu_find_pio_dev(vcpu, port); |
1631 | if (pio_dev) { | |
1632 | kernel_pio(pio_dev, vcpu, vcpu->pio_data); | |
1633 | complete_pio(vcpu); | |
1634 | return 1; | |
1635 | } | |
1636 | return 0; | |
1637 | } | |
1638 | EXPORT_SYMBOL_GPL(kvm_emulate_pio); | |
1639 | ||
1640 | int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in, | |
1641 | int size, unsigned long count, int down, | |
039576c0 AK |
1642 | gva_t address, int rep, unsigned port) |
1643 | { | |
1644 | unsigned now, in_page; | |
65619eb5 | 1645 | int i, ret = 0; |
039576c0 AK |
1646 | int nr_pages = 1; |
1647 | struct page *page; | |
74906345 | 1648 | struct kvm_io_device *pio_dev; |
039576c0 AK |
1649 | |
1650 | vcpu->run->exit_reason = KVM_EXIT_IO; | |
1651 | vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; | |
3090dd73 | 1652 | vcpu->run->io.size = vcpu->pio.size = size; |
039576c0 | 1653 | vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; |
3090dd73 LV |
1654 | vcpu->run->io.count = vcpu->pio.count = vcpu->pio.cur_count = count; |
1655 | vcpu->run->io.port = vcpu->pio.port = port; | |
039576c0 | 1656 | vcpu->pio.in = in; |
3090dd73 | 1657 | vcpu->pio.string = 1; |
039576c0 AK |
1658 | vcpu->pio.down = down; |
1659 | vcpu->pio.guest_page_offset = offset_in_page(address); | |
1660 | vcpu->pio.rep = rep; | |
1661 | ||
039576c0 | 1662 | if (!count) { |
cbdd1bea | 1663 | kvm_x86_ops->skip_emulated_instruction(vcpu); |
039576c0 AK |
1664 | return 1; |
1665 | } | |
1666 | ||
039576c0 AK |
1667 | if (!down) |
1668 | in_page = PAGE_SIZE - offset_in_page(address); | |
1669 | else | |
1670 | in_page = offset_in_page(address) + size; | |
1671 | now = min(count, (unsigned long)in_page / size); | |
1672 | if (!now) { | |
1673 | /* | |
1674 | * String I/O straddles page boundary. Pin two guest pages | |
1675 | * so that we satisfy atomicity constraints. Do just one | |
1676 | * transaction to avoid complexity. | |
1677 | */ | |
1678 | nr_pages = 2; | |
1679 | now = 1; | |
1680 | } | |
1681 | if (down) { | |
1682 | /* | |
1683 | * String I/O in reverse. Yuck. Kill the guest, fix later. | |
1684 | */ | |
f0242478 | 1685 | pr_unimpl(vcpu, "guest string pio down\n"); |
039576c0 AK |
1686 | inject_gp(vcpu); |
1687 | return 1; | |
1688 | } | |
1689 | vcpu->run->io.count = now; | |
1690 | vcpu->pio.cur_count = now; | |
1691 | ||
0967b7bf AK |
1692 | if (vcpu->pio.cur_count == vcpu->pio.count) |
1693 | kvm_x86_ops->skip_emulated_instruction(vcpu); | |
1694 | ||
039576c0 | 1695 | for (i = 0; i < nr_pages; ++i) { |
11ec2804 | 1696 | mutex_lock(&vcpu->kvm->lock); |
039576c0 | 1697 | page = gva_to_page(vcpu, address + i * PAGE_SIZE); |
039576c0 | 1698 | vcpu->pio.guest_pages[i] = page; |
11ec2804 | 1699 | mutex_unlock(&vcpu->kvm->lock); |
039576c0 AK |
1700 | if (!page) { |
1701 | inject_gp(vcpu); | |
1702 | free_pio_guest_pages(vcpu); | |
1703 | return 1; | |
1704 | } | |
1705 | } | |
1706 | ||
3090dd73 | 1707 | pio_dev = vcpu_find_pio_dev(vcpu, port); |
65619eb5 ED |
1708 | if (!vcpu->pio.in) { |
1709 | /* string PIO write */ | |
1710 | ret = pio_copy_data(vcpu); | |
1711 | if (ret >= 0 && pio_dev) { | |
1712 | pio_string_write(pio_dev, vcpu); | |
1713 | complete_pio(vcpu); | |
1714 | if (vcpu->pio.count == 0) | |
1715 | ret = 1; | |
1716 | } | |
1717 | } else if (pio_dev) | |
f0242478 | 1718 | pr_unimpl(vcpu, "no string pio read support yet, " |
65619eb5 ED |
1719 | "port %x size %d count %ld\n", |
1720 | port, size, count); | |
1721 | ||
1722 | return ret; | |
039576c0 | 1723 | } |
3090dd73 | 1724 | EXPORT_SYMBOL_GPL(kvm_emulate_pio_string); |
039576c0 | 1725 | |
04d2cc77 AK |
1726 | /* |
1727 | * Check if userspace requested an interrupt window, and that the | |
1728 | * interrupt window is open. | |
1729 | * | |
1730 | * No need to exit to userspace if we already have an interrupt queued. | |
1731 | */ | |
1732 | static int dm_request_for_irq_injection(struct kvm_vcpu *vcpu, | |
1733 | struct kvm_run *kvm_run) | |
1734 | { | |
1735 | return (!vcpu->irq_summary && | |
1736 | kvm_run->request_interrupt_window && | |
1737 | vcpu->interrupt_window_open && | |
1738 | (kvm_x86_ops->get_rflags(vcpu) & X86_EFLAGS_IF)); | |
1739 | } | |
1740 | ||
1741 | static void post_kvm_run_save(struct kvm_vcpu *vcpu, | |
1742 | struct kvm_run *kvm_run) | |
1743 | { | |
1744 | kvm_run->if_flag = (kvm_x86_ops->get_rflags(vcpu) & X86_EFLAGS_IF) != 0; | |
1745 | kvm_run->cr8 = get_cr8(vcpu); | |
1746 | kvm_run->apic_base = kvm_get_apic_base(vcpu); | |
1747 | if (irqchip_in_kernel(vcpu->kvm)) | |
1748 | kvm_run->ready_for_interrupt_injection = 1; | |
1749 | else | |
1750 | kvm_run->ready_for_interrupt_injection = | |
1751 | (vcpu->interrupt_window_open && | |
1752 | vcpu->irq_summary == 0); | |
1753 | } | |
1754 | ||
1755 | static int __vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) | |
1756 | { | |
1757 | int r; | |
1758 | ||
1759 | if (unlikely(vcpu->mp_state == VCPU_MP_STATE_SIPI_RECEIVED)) { | |
d77c26fc | 1760 | pr_debug("vcpu %d received sipi with vector # %x\n", |
04d2cc77 AK |
1761 | vcpu->vcpu_id, vcpu->sipi_vector); |
1762 | kvm_lapic_reset(vcpu); | |
e00c8cf2 AK |
1763 | r = kvm_x86_ops->vcpu_reset(vcpu); |
1764 | if (r) | |
1765 | return r; | |
04d2cc77 AK |
1766 | vcpu->mp_state = VCPU_MP_STATE_RUNNABLE; |
1767 | } | |
1768 | ||
1769 | preempted: | |
1770 | if (vcpu->guest_debug.enabled) | |
1771 | kvm_x86_ops->guest_debug_pre(vcpu); | |
1772 | ||
1773 | again: | |
1774 | r = kvm_mmu_reload(vcpu); | |
1775 | if (unlikely(r)) | |
1776 | goto out; | |
1777 | ||
ab6ef34b AK |
1778 | kvm_inject_pending_timer_irqs(vcpu); |
1779 | ||
04d2cc77 AK |
1780 | preempt_disable(); |
1781 | ||
1782 | kvm_x86_ops->prepare_guest_switch(vcpu); | |
1783 | kvm_load_guest_fpu(vcpu); | |
1784 | ||
1785 | local_irq_disable(); | |
1786 | ||
1787 | if (signal_pending(current)) { | |
1788 | local_irq_enable(); | |
1789 | preempt_enable(); | |
1790 | r = -EINTR; | |
1791 | kvm_run->exit_reason = KVM_EXIT_INTR; | |
1792 | ++vcpu->stat.signal_exits; | |
1793 | goto out; | |
1794 | } | |
1795 | ||
1796 | if (irqchip_in_kernel(vcpu->kvm)) | |
1797 | kvm_x86_ops->inject_pending_irq(vcpu); | |
1798 | else if (!vcpu->mmio_read_completed) | |
1799 | kvm_x86_ops->inject_pending_vectors(vcpu, kvm_run); | |
1800 | ||
1801 | vcpu->guest_mode = 1; | |
d172fcd3 | 1802 | kvm_guest_enter(); |
04d2cc77 AK |
1803 | |
1804 | if (vcpu->requests) | |
3176bc3e | 1805 | if (test_and_clear_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests)) |
04d2cc77 AK |
1806 | kvm_x86_ops->tlb_flush(vcpu); |
1807 | ||
1808 | kvm_x86_ops->run(vcpu, kvm_run); | |
1809 | ||
1810 | vcpu->guest_mode = 0; | |
1811 | local_irq_enable(); | |
1812 | ||
1813 | ++vcpu->stat.exits; | |
1814 | ||
0552f73b LV |
1815 | /* |
1816 | * We must have an instruction between local_irq_enable() and | |
1817 | * kvm_guest_exit(), so the timer interrupt isn't delayed by | |
1818 | * the interrupt shadow. The stat.exits increment will do nicely. | |
1819 | * But we need to prevent reordering, hence this barrier(): | |
1820 | */ | |
1821 | barrier(); | |
1822 | ||
1823 | kvm_guest_exit(); | |
1824 | ||
04d2cc77 AK |
1825 | preempt_enable(); |
1826 | ||
1827 | /* | |
1828 | * Profile KVM exit RIPs: | |
1829 | */ | |
1830 | if (unlikely(prof_on == KVM_PROFILING)) { | |
1831 | kvm_x86_ops->cache_regs(vcpu); | |
1832 | profile_hit(KVM_PROFILING, (void *)vcpu->rip); | |
1833 | } | |
1834 | ||
1835 | r = kvm_x86_ops->handle_exit(kvm_run, vcpu); | |
1836 | ||
1837 | if (r > 0) { | |
1838 | if (dm_request_for_irq_injection(vcpu, kvm_run)) { | |
1839 | r = -EINTR; | |
1840 | kvm_run->exit_reason = KVM_EXIT_INTR; | |
1841 | ++vcpu->stat.request_irq_exits; | |
1842 | goto out; | |
1843 | } | |
1844 | if (!need_resched()) { | |
1845 | ++vcpu->stat.light_exits; | |
1846 | goto again; | |
1847 | } | |
1848 | } | |
1849 | ||
1850 | out: | |
1851 | if (r > 0) { | |
1852 | kvm_resched(vcpu); | |
1853 | goto preempted; | |
1854 | } | |
1855 | ||
1856 | post_kvm_run_save(vcpu, kvm_run); | |
1857 | ||
1858 | return r; | |
1859 | } | |
1860 | ||
1861 | ||
bccf2150 | 1862 | static int kvm_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) |
6aa8b732 | 1863 | { |
6aa8b732 | 1864 | int r; |
1961d276 | 1865 | sigset_t sigsaved; |
6aa8b732 | 1866 | |
bccf2150 | 1867 | vcpu_load(vcpu); |
6aa8b732 | 1868 | |
c5ec1534 HQ |
1869 | if (unlikely(vcpu->mp_state == VCPU_MP_STATE_UNINITIALIZED)) { |
1870 | kvm_vcpu_block(vcpu); | |
1871 | vcpu_put(vcpu); | |
1872 | return -EAGAIN; | |
1873 | } | |
1874 | ||
1961d276 AK |
1875 | if (vcpu->sigset_active) |
1876 | sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); | |
1877 | ||
54810342 | 1878 | /* re-sync apic's tpr */ |
5cd4f6fd HQ |
1879 | if (!irqchip_in_kernel(vcpu->kvm)) |
1880 | set_cr8(vcpu, kvm_run->cr8); | |
54810342 | 1881 | |
02c83209 AK |
1882 | if (vcpu->pio.cur_count) { |
1883 | r = complete_pio(vcpu); | |
1884 | if (r) | |
1885 | goto out; | |
1886 | } | |
34c16eec | 1887 | #if CONFIG_HAS_IOMEM |
02c83209 AK |
1888 | if (vcpu->mmio_needed) { |
1889 | memcpy(vcpu->mmio_data, kvm_run->mmio.data, 8); | |
1890 | vcpu->mmio_read_completed = 1; | |
1891 | vcpu->mmio_needed = 0; | |
1892 | r = emulate_instruction(vcpu, kvm_run, | |
3427318f | 1893 | vcpu->mmio_fault_cr2, 0, 1); |
02c83209 AK |
1894 | if (r == EMULATE_DO_MMIO) { |
1895 | /* | |
1896 | * Read-modify-write. Back to userspace. | |
1897 | */ | |
02c83209 AK |
1898 | r = 0; |
1899 | goto out; | |
46fc1477 | 1900 | } |
6aa8b732 | 1901 | } |
34c16eec | 1902 | #endif |
8eb7d334 | 1903 | if (kvm_run->exit_reason == KVM_EXIT_HYPERCALL) { |
cbdd1bea | 1904 | kvm_x86_ops->cache_regs(vcpu); |
b4e63f56 | 1905 | vcpu->regs[VCPU_REGS_RAX] = kvm_run->hypercall.ret; |
cbdd1bea | 1906 | kvm_x86_ops->decache_regs(vcpu); |
b4e63f56 AK |
1907 | } |
1908 | ||
04d2cc77 | 1909 | r = __vcpu_run(vcpu, kvm_run); |
6aa8b732 | 1910 | |
039576c0 | 1911 | out: |
1961d276 AK |
1912 | if (vcpu->sigset_active) |
1913 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
1914 | ||
6aa8b732 AK |
1915 | vcpu_put(vcpu); |
1916 | return r; | |
1917 | } | |
1918 | ||
bccf2150 AK |
1919 | static int kvm_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, |
1920 | struct kvm_regs *regs) | |
6aa8b732 | 1921 | { |
bccf2150 | 1922 | vcpu_load(vcpu); |
6aa8b732 | 1923 | |
cbdd1bea | 1924 | kvm_x86_ops->cache_regs(vcpu); |
6aa8b732 AK |
1925 | |
1926 | regs->rax = vcpu->regs[VCPU_REGS_RAX]; | |
1927 | regs->rbx = vcpu->regs[VCPU_REGS_RBX]; | |
1928 | regs->rcx = vcpu->regs[VCPU_REGS_RCX]; | |
1929 | regs->rdx = vcpu->regs[VCPU_REGS_RDX]; | |
1930 | regs->rsi = vcpu->regs[VCPU_REGS_RSI]; | |
1931 | regs->rdi = vcpu->regs[VCPU_REGS_RDI]; | |
1932 | regs->rsp = vcpu->regs[VCPU_REGS_RSP]; | |
1933 | regs->rbp = vcpu->regs[VCPU_REGS_RBP]; | |
05b3e0c2 | 1934 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
1935 | regs->r8 = vcpu->regs[VCPU_REGS_R8]; |
1936 | regs->r9 = vcpu->regs[VCPU_REGS_R9]; | |
1937 | regs->r10 = vcpu->regs[VCPU_REGS_R10]; | |
1938 | regs->r11 = vcpu->regs[VCPU_REGS_R11]; | |
1939 | regs->r12 = vcpu->regs[VCPU_REGS_R12]; | |
1940 | regs->r13 = vcpu->regs[VCPU_REGS_R13]; | |
1941 | regs->r14 = vcpu->regs[VCPU_REGS_R14]; | |
1942 | regs->r15 = vcpu->regs[VCPU_REGS_R15]; | |
1943 | #endif | |
1944 | ||
1945 | regs->rip = vcpu->rip; | |
cbdd1bea | 1946 | regs->rflags = kvm_x86_ops->get_rflags(vcpu); |
6aa8b732 AK |
1947 | |
1948 | /* | |
1949 | * Don't leak debug flags in case they were set for guest debugging | |
1950 | */ | |
1951 | if (vcpu->guest_debug.enabled && vcpu->guest_debug.singlestep) | |
1952 | regs->rflags &= ~(X86_EFLAGS_TF | X86_EFLAGS_RF); | |
1953 | ||
1954 | vcpu_put(vcpu); | |
1955 | ||
1956 | return 0; | |
1957 | } | |
1958 | ||
bccf2150 AK |
1959 | static int kvm_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, |
1960 | struct kvm_regs *regs) | |
6aa8b732 | 1961 | { |
bccf2150 | 1962 | vcpu_load(vcpu); |
6aa8b732 AK |
1963 | |
1964 | vcpu->regs[VCPU_REGS_RAX] = regs->rax; | |
1965 | vcpu->regs[VCPU_REGS_RBX] = regs->rbx; | |
1966 | vcpu->regs[VCPU_REGS_RCX] = regs->rcx; | |
1967 | vcpu->regs[VCPU_REGS_RDX] = regs->rdx; | |
1968 | vcpu->regs[VCPU_REGS_RSI] = regs->rsi; | |
1969 | vcpu->regs[VCPU_REGS_RDI] = regs->rdi; | |
1970 | vcpu->regs[VCPU_REGS_RSP] = regs->rsp; | |
1971 | vcpu->regs[VCPU_REGS_RBP] = regs->rbp; | |
05b3e0c2 | 1972 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
1973 | vcpu->regs[VCPU_REGS_R8] = regs->r8; |
1974 | vcpu->regs[VCPU_REGS_R9] = regs->r9; | |
1975 | vcpu->regs[VCPU_REGS_R10] = regs->r10; | |
1976 | vcpu->regs[VCPU_REGS_R11] = regs->r11; | |
1977 | vcpu->regs[VCPU_REGS_R12] = regs->r12; | |
1978 | vcpu->regs[VCPU_REGS_R13] = regs->r13; | |
1979 | vcpu->regs[VCPU_REGS_R14] = regs->r14; | |
1980 | vcpu->regs[VCPU_REGS_R15] = regs->r15; | |
1981 | #endif | |
1982 | ||
1983 | vcpu->rip = regs->rip; | |
cbdd1bea | 1984 | kvm_x86_ops->set_rflags(vcpu, regs->rflags); |
6aa8b732 | 1985 | |
cbdd1bea | 1986 | kvm_x86_ops->decache_regs(vcpu); |
6aa8b732 AK |
1987 | |
1988 | vcpu_put(vcpu); | |
1989 | ||
1990 | return 0; | |
1991 | } | |
1992 | ||
1993 | static void get_segment(struct kvm_vcpu *vcpu, | |
1994 | struct kvm_segment *var, int seg) | |
1995 | { | |
cbdd1bea | 1996 | return kvm_x86_ops->get_segment(vcpu, var, seg); |
6aa8b732 AK |
1997 | } |
1998 | ||
bccf2150 AK |
1999 | static int kvm_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, |
2000 | struct kvm_sregs *sregs) | |
6aa8b732 | 2001 | { |
6aa8b732 | 2002 | struct descriptor_table dt; |
2a8067f1 | 2003 | int pending_vec; |
6aa8b732 | 2004 | |
bccf2150 | 2005 | vcpu_load(vcpu); |
6aa8b732 AK |
2006 | |
2007 | get_segment(vcpu, &sregs->cs, VCPU_SREG_CS); | |
2008 | get_segment(vcpu, &sregs->ds, VCPU_SREG_DS); | |
2009 | get_segment(vcpu, &sregs->es, VCPU_SREG_ES); | |
2010 | get_segment(vcpu, &sregs->fs, VCPU_SREG_FS); | |
2011 | get_segment(vcpu, &sregs->gs, VCPU_SREG_GS); | |
2012 | get_segment(vcpu, &sregs->ss, VCPU_SREG_SS); | |
2013 | ||
2014 | get_segment(vcpu, &sregs->tr, VCPU_SREG_TR); | |
2015 | get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); | |
2016 | ||
cbdd1bea | 2017 | kvm_x86_ops->get_idt(vcpu, &dt); |
6aa8b732 AK |
2018 | sregs->idt.limit = dt.limit; |
2019 | sregs->idt.base = dt.base; | |
cbdd1bea | 2020 | kvm_x86_ops->get_gdt(vcpu, &dt); |
6aa8b732 AK |
2021 | sregs->gdt.limit = dt.limit; |
2022 | sregs->gdt.base = dt.base; | |
2023 | ||
cbdd1bea | 2024 | kvm_x86_ops->decache_cr4_guest_bits(vcpu); |
6aa8b732 AK |
2025 | sregs->cr0 = vcpu->cr0; |
2026 | sregs->cr2 = vcpu->cr2; | |
2027 | sregs->cr3 = vcpu->cr3; | |
2028 | sregs->cr4 = vcpu->cr4; | |
7017fc3d | 2029 | sregs->cr8 = get_cr8(vcpu); |
6aa8b732 | 2030 | sregs->efer = vcpu->shadow_efer; |
7017fc3d | 2031 | sregs->apic_base = kvm_get_apic_base(vcpu); |
6aa8b732 | 2032 | |
2a8067f1 | 2033 | if (irqchip_in_kernel(vcpu->kvm)) { |
c52fb35a HQ |
2034 | memset(sregs->interrupt_bitmap, 0, |
2035 | sizeof sregs->interrupt_bitmap); | |
cbdd1bea | 2036 | pending_vec = kvm_x86_ops->get_irq(vcpu); |
2a8067f1 | 2037 | if (pending_vec >= 0) |
d77c26fc MD |
2038 | set_bit(pending_vec, |
2039 | (unsigned long *)sregs->interrupt_bitmap); | |
2a8067f1 | 2040 | } else |
c52fb35a HQ |
2041 | memcpy(sregs->interrupt_bitmap, vcpu->irq_pending, |
2042 | sizeof sregs->interrupt_bitmap); | |
6aa8b732 AK |
2043 | |
2044 | vcpu_put(vcpu); | |
2045 | ||
2046 | return 0; | |
2047 | } | |
2048 | ||
2049 | static void set_segment(struct kvm_vcpu *vcpu, | |
2050 | struct kvm_segment *var, int seg) | |
2051 | { | |
cbdd1bea | 2052 | return kvm_x86_ops->set_segment(vcpu, var, seg); |
6aa8b732 AK |
2053 | } |
2054 | ||
bccf2150 AK |
2055 | static int kvm_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, |
2056 | struct kvm_sregs *sregs) | |
6aa8b732 | 2057 | { |
6aa8b732 | 2058 | int mmu_reset_needed = 0; |
2a8067f1 | 2059 | int i, pending_vec, max_bits; |
6aa8b732 AK |
2060 | struct descriptor_table dt; |
2061 | ||
bccf2150 | 2062 | vcpu_load(vcpu); |
6aa8b732 | 2063 | |
6aa8b732 AK |
2064 | dt.limit = sregs->idt.limit; |
2065 | dt.base = sregs->idt.base; | |
cbdd1bea | 2066 | kvm_x86_ops->set_idt(vcpu, &dt); |
6aa8b732 AK |
2067 | dt.limit = sregs->gdt.limit; |
2068 | dt.base = sregs->gdt.base; | |
cbdd1bea | 2069 | kvm_x86_ops->set_gdt(vcpu, &dt); |
6aa8b732 AK |
2070 | |
2071 | vcpu->cr2 = sregs->cr2; | |
2072 | mmu_reset_needed |= vcpu->cr3 != sregs->cr3; | |
2073 | vcpu->cr3 = sregs->cr3; | |
2074 | ||
7017fc3d | 2075 | set_cr8(vcpu, sregs->cr8); |
6aa8b732 AK |
2076 | |
2077 | mmu_reset_needed |= vcpu->shadow_efer != sregs->efer; | |
05b3e0c2 | 2078 | #ifdef CONFIG_X86_64 |
cbdd1bea | 2079 | kvm_x86_ops->set_efer(vcpu, sregs->efer); |
6aa8b732 | 2080 | #endif |
7017fc3d | 2081 | kvm_set_apic_base(vcpu, sregs->apic_base); |
6aa8b732 | 2082 | |
cbdd1bea | 2083 | kvm_x86_ops->decache_cr4_guest_bits(vcpu); |
399badf3 | 2084 | |
6aa8b732 | 2085 | mmu_reset_needed |= vcpu->cr0 != sregs->cr0; |
81f50e3b | 2086 | vcpu->cr0 = sregs->cr0; |
cbdd1bea | 2087 | kvm_x86_ops->set_cr0(vcpu, sregs->cr0); |
6aa8b732 AK |
2088 | |
2089 | mmu_reset_needed |= vcpu->cr4 != sregs->cr4; | |
cbdd1bea | 2090 | kvm_x86_ops->set_cr4(vcpu, sregs->cr4); |
1b0973bd AK |
2091 | if (!is_long_mode(vcpu) && is_pae(vcpu)) |
2092 | load_pdptrs(vcpu, vcpu->cr3); | |
6aa8b732 AK |
2093 | |
2094 | if (mmu_reset_needed) | |
2095 | kvm_mmu_reset_context(vcpu); | |
2096 | ||
c52fb35a HQ |
2097 | if (!irqchip_in_kernel(vcpu->kvm)) { |
2098 | memcpy(vcpu->irq_pending, sregs->interrupt_bitmap, | |
2099 | sizeof vcpu->irq_pending); | |
2100 | vcpu->irq_summary = 0; | |
2101 | for (i = 0; i < ARRAY_SIZE(vcpu->irq_pending); ++i) | |
2102 | if (vcpu->irq_pending[i]) | |
2103 | __set_bit(i, &vcpu->irq_summary); | |
2a8067f1 ED |
2104 | } else { |
2105 | max_bits = (sizeof sregs->interrupt_bitmap) << 3; | |
2106 | pending_vec = find_first_bit( | |
2107 | (const unsigned long *)sregs->interrupt_bitmap, | |
2108 | max_bits); | |
2109 | /* Only pending external irq is handled here */ | |
2110 | if (pending_vec < max_bits) { | |
cbdd1bea | 2111 | kvm_x86_ops->set_irq(vcpu, pending_vec); |
d77c26fc MD |
2112 | pr_debug("Set back pending irq %d\n", |
2113 | pending_vec); | |
2a8067f1 | 2114 | } |
c52fb35a | 2115 | } |
6aa8b732 | 2116 | |
024aa1c0 AK |
2117 | set_segment(vcpu, &sregs->cs, VCPU_SREG_CS); |
2118 | set_segment(vcpu, &sregs->ds, VCPU_SREG_DS); | |
2119 | set_segment(vcpu, &sregs->es, VCPU_SREG_ES); | |
2120 | set_segment(vcpu, &sregs->fs, VCPU_SREG_FS); | |
2121 | set_segment(vcpu, &sregs->gs, VCPU_SREG_GS); | |
2122 | set_segment(vcpu, &sregs->ss, VCPU_SREG_SS); | |
2123 | ||
2124 | set_segment(vcpu, &sregs->tr, VCPU_SREG_TR); | |
2125 | set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); | |
2126 | ||
6aa8b732 AK |
2127 | vcpu_put(vcpu); |
2128 | ||
2129 | return 0; | |
2130 | } | |
2131 | ||
1747fb71 RR |
2132 | void kvm_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l) |
2133 | { | |
2134 | struct kvm_segment cs; | |
2135 | ||
2136 | get_segment(vcpu, &cs, VCPU_SREG_CS); | |
2137 | *db = cs.db; | |
2138 | *l = cs.l; | |
2139 | } | |
2140 | EXPORT_SYMBOL_GPL(kvm_get_cs_db_l_bits); | |
2141 | ||
6aa8b732 AK |
2142 | /* |
2143 | * Translate a guest virtual address to a guest physical address. | |
2144 | */ | |
bccf2150 AK |
2145 | static int kvm_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, |
2146 | struct kvm_translation *tr) | |
6aa8b732 AK |
2147 | { |
2148 | unsigned long vaddr = tr->linear_address; | |
6aa8b732 AK |
2149 | gpa_t gpa; |
2150 | ||
bccf2150 | 2151 | vcpu_load(vcpu); |
11ec2804 | 2152 | mutex_lock(&vcpu->kvm->lock); |
6aa8b732 AK |
2153 | gpa = vcpu->mmu.gva_to_gpa(vcpu, vaddr); |
2154 | tr->physical_address = gpa; | |
2155 | tr->valid = gpa != UNMAPPED_GVA; | |
2156 | tr->writeable = 1; | |
2157 | tr->usermode = 0; | |
11ec2804 | 2158 | mutex_unlock(&vcpu->kvm->lock); |
6aa8b732 AK |
2159 | vcpu_put(vcpu); |
2160 | ||
2161 | return 0; | |
2162 | } | |
2163 | ||
bccf2150 AK |
2164 | static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, |
2165 | struct kvm_interrupt *irq) | |
6aa8b732 | 2166 | { |
6aa8b732 AK |
2167 | if (irq->irq < 0 || irq->irq >= 256) |
2168 | return -EINVAL; | |
97222cc8 ED |
2169 | if (irqchip_in_kernel(vcpu->kvm)) |
2170 | return -ENXIO; | |
bccf2150 | 2171 | vcpu_load(vcpu); |
6aa8b732 AK |
2172 | |
2173 | set_bit(irq->irq, vcpu->irq_pending); | |
2174 | set_bit(irq->irq / BITS_PER_LONG, &vcpu->irq_summary); | |
2175 | ||
2176 | vcpu_put(vcpu); | |
2177 | ||
2178 | return 0; | |
2179 | } | |
2180 | ||
bccf2150 AK |
2181 | static int kvm_vcpu_ioctl_debug_guest(struct kvm_vcpu *vcpu, |
2182 | struct kvm_debug_guest *dbg) | |
6aa8b732 | 2183 | { |
6aa8b732 AK |
2184 | int r; |
2185 | ||
bccf2150 | 2186 | vcpu_load(vcpu); |
6aa8b732 | 2187 | |
cbdd1bea | 2188 | r = kvm_x86_ops->set_guest_debug(vcpu, dbg); |
6aa8b732 AK |
2189 | |
2190 | vcpu_put(vcpu); | |
2191 | ||
2192 | return r; | |
2193 | } | |
2194 | ||
9a2bb7f4 AK |
2195 | static struct page *kvm_vcpu_nopage(struct vm_area_struct *vma, |
2196 | unsigned long address, | |
2197 | int *type) | |
2198 | { | |
2199 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
2200 | unsigned long pgoff; | |
2201 | struct page *page; | |
2202 | ||
9a2bb7f4 | 2203 | pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
039576c0 AK |
2204 | if (pgoff == 0) |
2205 | page = virt_to_page(vcpu->run); | |
2206 | else if (pgoff == KVM_PIO_PAGE_OFFSET) | |
2207 | page = virt_to_page(vcpu->pio_data); | |
2208 | else | |
9a2bb7f4 | 2209 | return NOPAGE_SIGBUS; |
9a2bb7f4 | 2210 | get_page(page); |
cd0d9137 NAQ |
2211 | if (type != NULL) |
2212 | *type = VM_FAULT_MINOR; | |
2213 | ||
9a2bb7f4 AK |
2214 | return page; |
2215 | } | |
2216 | ||
2217 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
2218 | .nopage = kvm_vcpu_nopage, | |
2219 | }; | |
2220 | ||
2221 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
2222 | { | |
2223 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
2224 | return 0; | |
2225 | } | |
2226 | ||
bccf2150 AK |
2227 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
2228 | { | |
2229 | struct kvm_vcpu *vcpu = filp->private_data; | |
2230 | ||
2231 | fput(vcpu->kvm->filp); | |
2232 | return 0; | |
2233 | } | |
2234 | ||
2235 | static struct file_operations kvm_vcpu_fops = { | |
2236 | .release = kvm_vcpu_release, | |
2237 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
2238 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 2239 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
2240 | }; |
2241 | ||
2242 | /* | |
2243 | * Allocates an inode for the vcpu. | |
2244 | */ | |
2245 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
2246 | { | |
2247 | int fd, r; | |
2248 | struct inode *inode; | |
2249 | struct file *file; | |
2250 | ||
d6d28168 AK |
2251 | r = anon_inode_getfd(&fd, &inode, &file, |
2252 | "kvm-vcpu", &kvm_vcpu_fops, vcpu); | |
2253 | if (r) | |
2254 | return r; | |
bccf2150 | 2255 | atomic_inc(&vcpu->kvm->filp->f_count); |
bccf2150 | 2256 | return fd; |
bccf2150 AK |
2257 | } |
2258 | ||
c5ea7660 AK |
2259 | /* |
2260 | * Creates some virtual cpus. Good luck creating more than one. | |
2261 | */ | |
2262 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
2263 | { | |
2264 | int r; | |
2265 | struct kvm_vcpu *vcpu; | |
2266 | ||
c5ea7660 | 2267 | if (!valid_vcpu(n)) |
fb3f0f51 | 2268 | return -EINVAL; |
c5ea7660 | 2269 | |
cbdd1bea | 2270 | vcpu = kvm_x86_ops->vcpu_create(kvm, n); |
fb3f0f51 RR |
2271 | if (IS_ERR(vcpu)) |
2272 | return PTR_ERR(vcpu); | |
c5ea7660 | 2273 | |
15ad7146 AK |
2274 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
2275 | ||
b114b080 RR |
2276 | /* We do fxsave: this must be aligned. */ |
2277 | BUG_ON((unsigned long)&vcpu->host_fx_image & 0xF); | |
2278 | ||
fb3f0f51 | 2279 | vcpu_load(vcpu); |
e00c8cf2 AK |
2280 | r = kvm_x86_ops->vcpu_reset(vcpu); |
2281 | if (r == 0) | |
2282 | r = kvm_mmu_setup(vcpu); | |
c5ea7660 | 2283 | vcpu_put(vcpu); |
c5ea7660 | 2284 | if (r < 0) |
fb3f0f51 RR |
2285 | goto free_vcpu; |
2286 | ||
11ec2804 | 2287 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
2288 | if (kvm->vcpus[n]) { |
2289 | r = -EEXIST; | |
11ec2804 | 2290 | mutex_unlock(&kvm->lock); |
fb3f0f51 RR |
2291 | goto mmu_unload; |
2292 | } | |
2293 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 2294 | mutex_unlock(&kvm->lock); |
c5ea7660 | 2295 | |
fb3f0f51 | 2296 | /* Now it's all set up, let userspace reach it */ |
bccf2150 AK |
2297 | r = create_vcpu_fd(vcpu); |
2298 | if (r < 0) | |
fb3f0f51 RR |
2299 | goto unlink; |
2300 | return r; | |
39c3b86e | 2301 | |
fb3f0f51 | 2302 | unlink: |
11ec2804 | 2303 | mutex_lock(&kvm->lock); |
fb3f0f51 | 2304 | kvm->vcpus[n] = NULL; |
11ec2804 | 2305 | mutex_unlock(&kvm->lock); |
a2fa3e9f | 2306 | |
fb3f0f51 RR |
2307 | mmu_unload: |
2308 | vcpu_load(vcpu); | |
2309 | kvm_mmu_unload(vcpu); | |
2310 | vcpu_put(vcpu); | |
c5ea7660 | 2311 | |
fb3f0f51 | 2312 | free_vcpu: |
cbdd1bea | 2313 | kvm_x86_ops->vcpu_free(vcpu); |
c5ea7660 AK |
2314 | return r; |
2315 | } | |
2316 | ||
1961d276 AK |
2317 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
2318 | { | |
2319 | if (sigset) { | |
2320 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
2321 | vcpu->sigset_active = 1; | |
2322 | vcpu->sigset = *sigset; | |
2323 | } else | |
2324 | vcpu->sigset_active = 0; | |
2325 | return 0; | |
2326 | } | |
2327 | ||
b8836737 AK |
2328 | /* |
2329 | * fxsave fpu state. Taken from x86_64/processor.h. To be killed when | |
2330 | * we have asm/x86/processor.h | |
2331 | */ | |
2332 | struct fxsave { | |
2333 | u16 cwd; | |
2334 | u16 swd; | |
2335 | u16 twd; | |
2336 | u16 fop; | |
2337 | u64 rip; | |
2338 | u64 rdp; | |
2339 | u32 mxcsr; | |
2340 | u32 mxcsr_mask; | |
2341 | u32 st_space[32]; /* 8*16 bytes for each FP-reg = 128 bytes */ | |
2342 | #ifdef CONFIG_X86_64 | |
2343 | u32 xmm_space[64]; /* 16*16 bytes for each XMM-reg = 256 bytes */ | |
2344 | #else | |
2345 | u32 xmm_space[32]; /* 8*16 bytes for each XMM-reg = 128 bytes */ | |
2346 | #endif | |
2347 | }; | |
2348 | ||
2349 | static int kvm_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
2350 | { | |
b114b080 | 2351 | struct fxsave *fxsave = (struct fxsave *)&vcpu->guest_fx_image; |
b8836737 AK |
2352 | |
2353 | vcpu_load(vcpu); | |
2354 | ||
2355 | memcpy(fpu->fpr, fxsave->st_space, 128); | |
2356 | fpu->fcw = fxsave->cwd; | |
2357 | fpu->fsw = fxsave->swd; | |
2358 | fpu->ftwx = fxsave->twd; | |
2359 | fpu->last_opcode = fxsave->fop; | |
2360 | fpu->last_ip = fxsave->rip; | |
2361 | fpu->last_dp = fxsave->rdp; | |
2362 | memcpy(fpu->xmm, fxsave->xmm_space, sizeof fxsave->xmm_space); | |
2363 | ||
2364 | vcpu_put(vcpu); | |
2365 | ||
2366 | return 0; | |
2367 | } | |
2368 | ||
2369 | static int kvm_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
2370 | { | |
b114b080 | 2371 | struct fxsave *fxsave = (struct fxsave *)&vcpu->guest_fx_image; |
b8836737 AK |
2372 | |
2373 | vcpu_load(vcpu); | |
2374 | ||
2375 | memcpy(fxsave->st_space, fpu->fpr, 128); | |
2376 | fxsave->cwd = fpu->fcw; | |
2377 | fxsave->swd = fpu->fsw; | |
2378 | fxsave->twd = fpu->ftwx; | |
2379 | fxsave->fop = fpu->last_opcode; | |
2380 | fxsave->rip = fpu->last_ip; | |
2381 | fxsave->rdp = fpu->last_dp; | |
2382 | memcpy(fxsave->xmm_space, fpu->xmm, sizeof fxsave->xmm_space); | |
2383 | ||
2384 | vcpu_put(vcpu); | |
2385 | ||
2386 | return 0; | |
2387 | } | |
2388 | ||
bccf2150 AK |
2389 | static long kvm_vcpu_ioctl(struct file *filp, |
2390 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 2391 | { |
bccf2150 | 2392 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 2393 | void __user *argp = (void __user *)arg; |
313a3dc7 | 2394 | int r; |
6aa8b732 AK |
2395 | |
2396 | switch (ioctl) { | |
9a2bb7f4 | 2397 | case KVM_RUN: |
f0fe5108 AK |
2398 | r = -EINVAL; |
2399 | if (arg) | |
2400 | goto out; | |
9a2bb7f4 | 2401 | r = kvm_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 2402 | break; |
6aa8b732 AK |
2403 | case KVM_GET_REGS: { |
2404 | struct kvm_regs kvm_regs; | |
2405 | ||
bccf2150 AK |
2406 | memset(&kvm_regs, 0, sizeof kvm_regs); |
2407 | r = kvm_vcpu_ioctl_get_regs(vcpu, &kvm_regs); | |
6aa8b732 AK |
2408 | if (r) |
2409 | goto out; | |
2410 | r = -EFAULT; | |
2f366987 | 2411 | if (copy_to_user(argp, &kvm_regs, sizeof kvm_regs)) |
6aa8b732 AK |
2412 | goto out; |
2413 | r = 0; | |
2414 | break; | |
2415 | } | |
2416 | case KVM_SET_REGS: { | |
2417 | struct kvm_regs kvm_regs; | |
2418 | ||
2419 | r = -EFAULT; | |
2f366987 | 2420 | if (copy_from_user(&kvm_regs, argp, sizeof kvm_regs)) |
6aa8b732 | 2421 | goto out; |
bccf2150 | 2422 | r = kvm_vcpu_ioctl_set_regs(vcpu, &kvm_regs); |
6aa8b732 AK |
2423 | if (r) |
2424 | goto out; | |
2425 | r = 0; | |
2426 | break; | |
2427 | } | |
2428 | case KVM_GET_SREGS: { | |
2429 | struct kvm_sregs kvm_sregs; | |
2430 | ||
bccf2150 AK |
2431 | memset(&kvm_sregs, 0, sizeof kvm_sregs); |
2432 | r = kvm_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs); | |
6aa8b732 AK |
2433 | if (r) |
2434 | goto out; | |
2435 | r = -EFAULT; | |
2f366987 | 2436 | if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs)) |
6aa8b732 AK |
2437 | goto out; |
2438 | r = 0; | |
2439 | break; | |
2440 | } | |
2441 | case KVM_SET_SREGS: { | |
2442 | struct kvm_sregs kvm_sregs; | |
2443 | ||
2444 | r = -EFAULT; | |
2f366987 | 2445 | if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs)) |
6aa8b732 | 2446 | goto out; |
bccf2150 | 2447 | r = kvm_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
2448 | if (r) |
2449 | goto out; | |
2450 | r = 0; | |
2451 | break; | |
2452 | } | |
2453 | case KVM_TRANSLATE: { | |
2454 | struct kvm_translation tr; | |
2455 | ||
2456 | r = -EFAULT; | |
2f366987 | 2457 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 2458 | goto out; |
bccf2150 | 2459 | r = kvm_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
2460 | if (r) |
2461 | goto out; | |
2462 | r = -EFAULT; | |
2f366987 | 2463 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
2464 | goto out; |
2465 | r = 0; | |
2466 | break; | |
2467 | } | |
2468 | case KVM_INTERRUPT: { | |
2469 | struct kvm_interrupt irq; | |
2470 | ||
2471 | r = -EFAULT; | |
2f366987 | 2472 | if (copy_from_user(&irq, argp, sizeof irq)) |
6aa8b732 | 2473 | goto out; |
bccf2150 | 2474 | r = kvm_vcpu_ioctl_interrupt(vcpu, &irq); |
6aa8b732 AK |
2475 | if (r) |
2476 | goto out; | |
2477 | r = 0; | |
2478 | break; | |
2479 | } | |
2480 | case KVM_DEBUG_GUEST: { | |
2481 | struct kvm_debug_guest dbg; | |
2482 | ||
2483 | r = -EFAULT; | |
2f366987 | 2484 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 2485 | goto out; |
bccf2150 | 2486 | r = kvm_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
2487 | if (r) |
2488 | goto out; | |
2489 | r = 0; | |
2490 | break; | |
2491 | } | |
1961d276 AK |
2492 | case KVM_SET_SIGNAL_MASK: { |
2493 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
2494 | struct kvm_signal_mask kvm_sigmask; | |
2495 | sigset_t sigset, *p; | |
2496 | ||
2497 | p = NULL; | |
2498 | if (argp) { | |
2499 | r = -EFAULT; | |
2500 | if (copy_from_user(&kvm_sigmask, argp, | |
2501 | sizeof kvm_sigmask)) | |
2502 | goto out; | |
2503 | r = -EINVAL; | |
2504 | if (kvm_sigmask.len != sizeof sigset) | |
2505 | goto out; | |
2506 | r = -EFAULT; | |
2507 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
2508 | sizeof sigset)) | |
2509 | goto out; | |
2510 | p = &sigset; | |
2511 | } | |
2512 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
2513 | break; | |
2514 | } | |
b8836737 AK |
2515 | case KVM_GET_FPU: { |
2516 | struct kvm_fpu fpu; | |
2517 | ||
2518 | memset(&fpu, 0, sizeof fpu); | |
2519 | r = kvm_vcpu_ioctl_get_fpu(vcpu, &fpu); | |
2520 | if (r) | |
2521 | goto out; | |
2522 | r = -EFAULT; | |
2523 | if (copy_to_user(argp, &fpu, sizeof fpu)) | |
2524 | goto out; | |
2525 | r = 0; | |
2526 | break; | |
2527 | } | |
2528 | case KVM_SET_FPU: { | |
2529 | struct kvm_fpu fpu; | |
2530 | ||
2531 | r = -EFAULT; | |
2532 | if (copy_from_user(&fpu, argp, sizeof fpu)) | |
2533 | goto out; | |
2534 | r = kvm_vcpu_ioctl_set_fpu(vcpu, &fpu); | |
2535 | if (r) | |
2536 | goto out; | |
2537 | r = 0; | |
2538 | break; | |
2539 | } | |
bccf2150 | 2540 | default: |
313a3dc7 | 2541 | r = kvm_arch_vcpu_ioctl(filp, ioctl, arg); |
bccf2150 AK |
2542 | } |
2543 | out: | |
2544 | return r; | |
2545 | } | |
2546 | ||
2547 | static long kvm_vm_ioctl(struct file *filp, | |
2548 | unsigned int ioctl, unsigned long arg) | |
2549 | { | |
2550 | struct kvm *kvm = filp->private_data; | |
2551 | void __user *argp = (void __user *)arg; | |
1fe779f8 | 2552 | int r; |
bccf2150 AK |
2553 | |
2554 | switch (ioctl) { | |
2555 | case KVM_CREATE_VCPU: | |
2556 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
2557 | if (r < 0) | |
2558 | goto out; | |
2559 | break; | |
6fc138d2 IE |
2560 | case KVM_SET_USER_MEMORY_REGION: { |
2561 | struct kvm_userspace_memory_region kvm_userspace_mem; | |
2562 | ||
2563 | r = -EFAULT; | |
2564 | if (copy_from_user(&kvm_userspace_mem, argp, | |
2565 | sizeof kvm_userspace_mem)) | |
2566 | goto out; | |
2567 | ||
2568 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1); | |
6aa8b732 AK |
2569 | if (r) |
2570 | goto out; | |
2571 | break; | |
2572 | } | |
2573 | case KVM_GET_DIRTY_LOG: { | |
2574 | struct kvm_dirty_log log; | |
2575 | ||
2576 | r = -EFAULT; | |
2f366987 | 2577 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2578 | goto out; |
2c6f5df9 | 2579 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2580 | if (r) |
2581 | goto out; | |
2582 | break; | |
2583 | } | |
f17abe9a | 2584 | default: |
1fe779f8 | 2585 | r = kvm_arch_vm_ioctl(filp, ioctl, arg); |
f17abe9a AK |
2586 | } |
2587 | out: | |
2588 | return r; | |
2589 | } | |
2590 | ||
2591 | static struct page *kvm_vm_nopage(struct vm_area_struct *vma, | |
2592 | unsigned long address, | |
2593 | int *type) | |
2594 | { | |
2595 | struct kvm *kvm = vma->vm_file->private_data; | |
2596 | unsigned long pgoff; | |
f17abe9a AK |
2597 | struct page *page; |
2598 | ||
f17abe9a | 2599 | pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
e0d62c7f IE |
2600 | if (!kvm_is_visible_gfn(kvm, pgoff)) |
2601 | return NOPAGE_SIGBUS; | |
954bbbc2 | 2602 | page = gfn_to_page(kvm, pgoff); |
8a7ae055 IE |
2603 | if (is_error_page(page)) { |
2604 | kvm_release_page(page); | |
f17abe9a | 2605 | return NOPAGE_SIGBUS; |
8a7ae055 | 2606 | } |
cd0d9137 NAQ |
2607 | if (type != NULL) |
2608 | *type = VM_FAULT_MINOR; | |
2609 | ||
f17abe9a AK |
2610 | return page; |
2611 | } | |
2612 | ||
2613 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
2614 | .nopage = kvm_vm_nopage, | |
2615 | }; | |
2616 | ||
2617 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
2618 | { | |
2619 | vma->vm_ops = &kvm_vm_vm_ops; | |
2620 | return 0; | |
2621 | } | |
2622 | ||
2623 | static struct file_operations kvm_vm_fops = { | |
2624 | .release = kvm_vm_release, | |
2625 | .unlocked_ioctl = kvm_vm_ioctl, | |
2626 | .compat_ioctl = kvm_vm_ioctl, | |
2627 | .mmap = kvm_vm_mmap, | |
2628 | }; | |
2629 | ||
2630 | static int kvm_dev_ioctl_create_vm(void) | |
2631 | { | |
2632 | int fd, r; | |
2633 | struct inode *inode; | |
2634 | struct file *file; | |
2635 | struct kvm *kvm; | |
2636 | ||
f17abe9a | 2637 | kvm = kvm_create_vm(); |
d6d28168 AK |
2638 | if (IS_ERR(kvm)) |
2639 | return PTR_ERR(kvm); | |
2640 | r = anon_inode_getfd(&fd, &inode, &file, "kvm-vm", &kvm_vm_fops, kvm); | |
2641 | if (r) { | |
2642 | kvm_destroy_vm(kvm); | |
2643 | return r; | |
f17abe9a AK |
2644 | } |
2645 | ||
bccf2150 | 2646 | kvm->filp = file; |
f17abe9a | 2647 | |
f17abe9a | 2648 | return fd; |
f17abe9a AK |
2649 | } |
2650 | ||
2651 | static long kvm_dev_ioctl(struct file *filp, | |
2652 | unsigned int ioctl, unsigned long arg) | |
2653 | { | |
2654 | void __user *argp = (void __user *)arg; | |
07c45a36 | 2655 | long r = -EINVAL; |
f17abe9a AK |
2656 | |
2657 | switch (ioctl) { | |
2658 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
2659 | r = -EINVAL; |
2660 | if (arg) | |
2661 | goto out; | |
f17abe9a AK |
2662 | r = KVM_API_VERSION; |
2663 | break; | |
2664 | case KVM_CREATE_VM: | |
f0fe5108 AK |
2665 | r = -EINVAL; |
2666 | if (arg) | |
2667 | goto out; | |
f17abe9a AK |
2668 | r = kvm_dev_ioctl_create_vm(); |
2669 | break; | |
85f455f7 ED |
2670 | case KVM_CHECK_EXTENSION: { |
2671 | int ext = (long)argp; | |
2672 | ||
2673 | switch (ext) { | |
2674 | case KVM_CAP_IRQCHIP: | |
b6958ce4 | 2675 | case KVM_CAP_HLT: |
82ce2c96 | 2676 | case KVM_CAP_MMU_SHADOW_CACHE_CONTROL: |
6fc138d2 | 2677 | case KVM_CAP_USER_MEMORY: |
cbc94022 | 2678 | case KVM_CAP_SET_TSS_ADDR: |
85f455f7 ED |
2679 | r = 1; |
2680 | break; | |
2681 | default: | |
2682 | r = 0; | |
2683 | break; | |
2684 | } | |
5d308f45 | 2685 | break; |
85f455f7 | 2686 | } |
07c45a36 AK |
2687 | case KVM_GET_VCPU_MMAP_SIZE: |
2688 | r = -EINVAL; | |
2689 | if (arg) | |
2690 | goto out; | |
039576c0 | 2691 | r = 2 * PAGE_SIZE; |
07c45a36 | 2692 | break; |
6aa8b732 | 2693 | default: |
043405e1 | 2694 | return kvm_arch_dev_ioctl(filp, ioctl, arg); |
6aa8b732 AK |
2695 | } |
2696 | out: | |
2697 | return r; | |
2698 | } | |
2699 | ||
6aa8b732 | 2700 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
2701 | .unlocked_ioctl = kvm_dev_ioctl, |
2702 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
2703 | }; |
2704 | ||
2705 | static struct miscdevice kvm_dev = { | |
bbe4432e | 2706 | KVM_MINOR, |
6aa8b732 AK |
2707 | "kvm", |
2708 | &kvm_chardev_ops, | |
2709 | }; | |
2710 | ||
774c47f1 AK |
2711 | /* |
2712 | * Make sure that a cpu that is being hot-unplugged does not have any vcpus | |
2713 | * cached on it. | |
2714 | */ | |
2715 | static void decache_vcpus_on_cpu(int cpu) | |
2716 | { | |
2717 | struct kvm *vm; | |
2718 | struct kvm_vcpu *vcpu; | |
2719 | int i; | |
2720 | ||
2721 | spin_lock(&kvm_lock); | |
11ec2804 | 2722 | list_for_each_entry(vm, &vm_list, vm_list) |
774c47f1 | 2723 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
fb3f0f51 RR |
2724 | vcpu = vm->vcpus[i]; |
2725 | if (!vcpu) | |
2726 | continue; | |
774c47f1 AK |
2727 | /* |
2728 | * If the vcpu is locked, then it is running on some | |
2729 | * other cpu and therefore it is not cached on the | |
2730 | * cpu in question. | |
2731 | * | |
2732 | * If it's not locked, check the last cpu it executed | |
2733 | * on. | |
2734 | */ | |
2735 | if (mutex_trylock(&vcpu->mutex)) { | |
2736 | if (vcpu->cpu == cpu) { | |
cbdd1bea | 2737 | kvm_x86_ops->vcpu_decache(vcpu); |
774c47f1 AK |
2738 | vcpu->cpu = -1; |
2739 | } | |
2740 | mutex_unlock(&vcpu->mutex); | |
2741 | } | |
2742 | } | |
2743 | spin_unlock(&kvm_lock); | |
2744 | } | |
2745 | ||
1b6c0168 AK |
2746 | static void hardware_enable(void *junk) |
2747 | { | |
2748 | int cpu = raw_smp_processor_id(); | |
2749 | ||
2750 | if (cpu_isset(cpu, cpus_hardware_enabled)) | |
2751 | return; | |
2752 | cpu_set(cpu, cpus_hardware_enabled); | |
cbdd1bea | 2753 | kvm_x86_ops->hardware_enable(NULL); |
1b6c0168 AK |
2754 | } |
2755 | ||
2756 | static void hardware_disable(void *junk) | |
2757 | { | |
2758 | int cpu = raw_smp_processor_id(); | |
2759 | ||
2760 | if (!cpu_isset(cpu, cpus_hardware_enabled)) | |
2761 | return; | |
2762 | cpu_clear(cpu, cpus_hardware_enabled); | |
2763 | decache_vcpus_on_cpu(cpu); | |
cbdd1bea | 2764 | kvm_x86_ops->hardware_disable(NULL); |
1b6c0168 AK |
2765 | } |
2766 | ||
774c47f1 AK |
2767 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
2768 | void *v) | |
2769 | { | |
2770 | int cpu = (long)v; | |
2771 | ||
2772 | switch (val) { | |
cec9ad27 AK |
2773 | case CPU_DYING: |
2774 | case CPU_DYING_FROZEN: | |
6ec8a856 AK |
2775 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2776 | cpu); | |
2777 | hardware_disable(NULL); | |
2778 | break; | |
774c47f1 | 2779 | case CPU_UP_CANCELED: |
8bb78442 | 2780 | case CPU_UP_CANCELED_FROZEN: |
43934a38 JK |
2781 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
2782 | cpu); | |
1b6c0168 | 2783 | smp_call_function_single(cpu, hardware_disable, NULL, 0, 1); |
774c47f1 | 2784 | break; |
43934a38 | 2785 | case CPU_ONLINE: |
8bb78442 | 2786 | case CPU_ONLINE_FROZEN: |
43934a38 JK |
2787 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
2788 | cpu); | |
1b6c0168 | 2789 | smp_call_function_single(cpu, hardware_enable, NULL, 0, 1); |
774c47f1 AK |
2790 | break; |
2791 | } | |
2792 | return NOTIFY_OK; | |
2793 | } | |
2794 | ||
9a2b85c6 | 2795 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
d77c26fc | 2796 | void *v) |
9a2b85c6 RR |
2797 | { |
2798 | if (val == SYS_RESTART) { | |
2799 | /* | |
2800 | * Some (well, at least mine) BIOSes hang on reboot if | |
2801 | * in vmx root mode. | |
2802 | */ | |
2803 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
2804 | on_each_cpu(hardware_disable, NULL, 0, 1); | |
2805 | } | |
2806 | return NOTIFY_OK; | |
2807 | } | |
2808 | ||
2809 | static struct notifier_block kvm_reboot_notifier = { | |
2810 | .notifier_call = kvm_reboot, | |
2811 | .priority = 0, | |
2812 | }; | |
2813 | ||
2eeb2e94 GH |
2814 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
2815 | { | |
2816 | memset(bus, 0, sizeof(*bus)); | |
2817 | } | |
2818 | ||
2819 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
2820 | { | |
2821 | int i; | |
2822 | ||
2823 | for (i = 0; i < bus->dev_count; i++) { | |
2824 | struct kvm_io_device *pos = bus->devs[i]; | |
2825 | ||
2826 | kvm_iodevice_destructor(pos); | |
2827 | } | |
2828 | } | |
2829 | ||
2830 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr) | |
2831 | { | |
2832 | int i; | |
2833 | ||
2834 | for (i = 0; i < bus->dev_count; i++) { | |
2835 | struct kvm_io_device *pos = bus->devs[i]; | |
2836 | ||
2837 | if (pos->in_range(pos, addr)) | |
2838 | return pos; | |
2839 | } | |
2840 | ||
2841 | return NULL; | |
2842 | } | |
2843 | ||
2844 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
2845 | { | |
2846 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
2847 | ||
2848 | bus->devs[bus->dev_count++] = dev; | |
2849 | } | |
2850 | ||
774c47f1 AK |
2851 | static struct notifier_block kvm_cpu_notifier = { |
2852 | .notifier_call = kvm_cpu_hotplug, | |
2853 | .priority = 20, /* must be > scheduler priority */ | |
2854 | }; | |
2855 | ||
1165f5fe AK |
2856 | static u64 stat_get(void *_offset) |
2857 | { | |
2858 | unsigned offset = (long)_offset; | |
2859 | u64 total = 0; | |
2860 | struct kvm *kvm; | |
2861 | struct kvm_vcpu *vcpu; | |
2862 | int i; | |
2863 | ||
2864 | spin_lock(&kvm_lock); | |
2865 | list_for_each_entry(kvm, &vm_list, vm_list) | |
2866 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
2867 | vcpu = kvm->vcpus[i]; |
2868 | if (vcpu) | |
2869 | total += *(u32 *)((void *)vcpu + offset); | |
1165f5fe AK |
2870 | } |
2871 | spin_unlock(&kvm_lock); | |
2872 | return total; | |
2873 | } | |
2874 | ||
3dea7ca7 | 2875 | DEFINE_SIMPLE_ATTRIBUTE(stat_fops, stat_get, NULL, "%llu\n"); |
1165f5fe | 2876 | |
6aa8b732 AK |
2877 | static __init void kvm_init_debug(void) |
2878 | { | |
2879 | struct kvm_stats_debugfs_item *p; | |
2880 | ||
8b6d44c7 | 2881 | debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 2882 | for (p = debugfs_entries; p->name; ++p) |
1165f5fe AK |
2883 | p->dentry = debugfs_create_file(p->name, 0444, debugfs_dir, |
2884 | (void *)(long)p->offset, | |
2885 | &stat_fops); | |
6aa8b732 AK |
2886 | } |
2887 | ||
2888 | static void kvm_exit_debug(void) | |
2889 | { | |
2890 | struct kvm_stats_debugfs_item *p; | |
2891 | ||
2892 | for (p = debugfs_entries; p->name; ++p) | |
2893 | debugfs_remove(p->dentry); | |
2894 | debugfs_remove(debugfs_dir); | |
2895 | } | |
2896 | ||
59ae6c6b AK |
2897 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
2898 | { | |
4267c41a | 2899 | hardware_disable(NULL); |
59ae6c6b AK |
2900 | return 0; |
2901 | } | |
2902 | ||
2903 | static int kvm_resume(struct sys_device *dev) | |
2904 | { | |
4267c41a | 2905 | hardware_enable(NULL); |
59ae6c6b AK |
2906 | return 0; |
2907 | } | |
2908 | ||
2909 | static struct sysdev_class kvm_sysdev_class = { | |
af5ca3f4 | 2910 | .name = "kvm", |
59ae6c6b AK |
2911 | .suspend = kvm_suspend, |
2912 | .resume = kvm_resume, | |
2913 | }; | |
2914 | ||
2915 | static struct sys_device kvm_sysdev = { | |
2916 | .id = 0, | |
2917 | .cls = &kvm_sysdev_class, | |
2918 | }; | |
2919 | ||
cea7bb21 | 2920 | struct page *bad_page; |
6aa8b732 | 2921 | |
15ad7146 AK |
2922 | static inline |
2923 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
2924 | { | |
2925 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
2926 | } | |
2927 | ||
2928 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
2929 | { | |
2930 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2931 | ||
cbdd1bea | 2932 | kvm_x86_ops->vcpu_load(vcpu, cpu); |
15ad7146 AK |
2933 | } |
2934 | ||
2935 | static void kvm_sched_out(struct preempt_notifier *pn, | |
2936 | struct task_struct *next) | |
2937 | { | |
2938 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
2939 | ||
cbdd1bea | 2940 | kvm_x86_ops->vcpu_put(vcpu); |
15ad7146 AK |
2941 | } |
2942 | ||
cbdd1bea | 2943 | int kvm_init_x86(struct kvm_x86_ops *ops, unsigned int vcpu_size, |
c16f862d | 2944 | struct module *module) |
6aa8b732 AK |
2945 | { |
2946 | int r; | |
002c7f7c | 2947 | int cpu; |
6aa8b732 | 2948 | |
cbdd1bea | 2949 | if (kvm_x86_ops) { |
09db28b8 YI |
2950 | printk(KERN_ERR "kvm: already loaded the other module\n"); |
2951 | return -EEXIST; | |
2952 | } | |
2953 | ||
e097f35c | 2954 | if (!ops->cpu_has_kvm_support()) { |
6aa8b732 AK |
2955 | printk(KERN_ERR "kvm: no hardware support\n"); |
2956 | return -EOPNOTSUPP; | |
2957 | } | |
e097f35c | 2958 | if (ops->disabled_by_bios()) { |
6aa8b732 AK |
2959 | printk(KERN_ERR "kvm: disabled by bios\n"); |
2960 | return -EOPNOTSUPP; | |
2961 | } | |
2962 | ||
cbdd1bea | 2963 | kvm_x86_ops = ops; |
e097f35c | 2964 | |
cbdd1bea | 2965 | r = kvm_x86_ops->hardware_setup(); |
6aa8b732 | 2966 | if (r < 0) |
ca45aaae | 2967 | goto out; |
6aa8b732 | 2968 | |
002c7f7c YS |
2969 | for_each_online_cpu(cpu) { |
2970 | smp_call_function_single(cpu, | |
cbdd1bea | 2971 | kvm_x86_ops->check_processor_compatibility, |
002c7f7c YS |
2972 | &r, 0, 1); |
2973 | if (r < 0) | |
2974 | goto out_free_0; | |
2975 | } | |
2976 | ||
1b6c0168 | 2977 | on_each_cpu(hardware_enable, NULL, 0, 1); |
774c47f1 AK |
2978 | r = register_cpu_notifier(&kvm_cpu_notifier); |
2979 | if (r) | |
2980 | goto out_free_1; | |
6aa8b732 AK |
2981 | register_reboot_notifier(&kvm_reboot_notifier); |
2982 | ||
59ae6c6b AK |
2983 | r = sysdev_class_register(&kvm_sysdev_class); |
2984 | if (r) | |
2985 | goto out_free_2; | |
2986 | ||
2987 | r = sysdev_register(&kvm_sysdev); | |
2988 | if (r) | |
2989 | goto out_free_3; | |
2990 | ||
c16f862d RR |
2991 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
2992 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
2993 | __alignof__(struct kvm_vcpu), 0, 0); | |
2994 | if (!kvm_vcpu_cache) { | |
2995 | r = -ENOMEM; | |
2996 | goto out_free_4; | |
2997 | } | |
2998 | ||
6aa8b732 AK |
2999 | kvm_chardev_ops.owner = module; |
3000 | ||
3001 | r = misc_register(&kvm_dev); | |
3002 | if (r) { | |
d77c26fc | 3003 | printk(KERN_ERR "kvm: misc device register failed\n"); |
6aa8b732 AK |
3004 | goto out_free; |
3005 | } | |
3006 | ||
15ad7146 AK |
3007 | kvm_preempt_ops.sched_in = kvm_sched_in; |
3008 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
3009 | ||
c7addb90 AK |
3010 | kvm_mmu_set_nonpresent_ptes(0ull, 0ull); |
3011 | ||
3012 | return 0; | |
6aa8b732 AK |
3013 | |
3014 | out_free: | |
c16f862d RR |
3015 | kmem_cache_destroy(kvm_vcpu_cache); |
3016 | out_free_4: | |
59ae6c6b AK |
3017 | sysdev_unregister(&kvm_sysdev); |
3018 | out_free_3: | |
3019 | sysdev_class_unregister(&kvm_sysdev_class); | |
3020 | out_free_2: | |
6aa8b732 | 3021 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 AK |
3022 | unregister_cpu_notifier(&kvm_cpu_notifier); |
3023 | out_free_1: | |
1b6c0168 | 3024 | on_each_cpu(hardware_disable, NULL, 0, 1); |
002c7f7c | 3025 | out_free_0: |
cbdd1bea | 3026 | kvm_x86_ops->hardware_unsetup(); |
ca45aaae | 3027 | out: |
cbdd1bea | 3028 | kvm_x86_ops = NULL; |
6aa8b732 AK |
3029 | return r; |
3030 | } | |
d77c26fc | 3031 | EXPORT_SYMBOL_GPL(kvm_init_x86); |
6aa8b732 | 3032 | |
cbdd1bea | 3033 | void kvm_exit_x86(void) |
6aa8b732 AK |
3034 | { |
3035 | misc_deregister(&kvm_dev); | |
c16f862d | 3036 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
3037 | sysdev_unregister(&kvm_sysdev); |
3038 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 3039 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 3040 | unregister_cpu_notifier(&kvm_cpu_notifier); |
1b6c0168 | 3041 | on_each_cpu(hardware_disable, NULL, 0, 1); |
cbdd1bea CE |
3042 | kvm_x86_ops->hardware_unsetup(); |
3043 | kvm_x86_ops = NULL; | |
6aa8b732 | 3044 | } |
d77c26fc | 3045 | EXPORT_SYMBOL_GPL(kvm_exit_x86); |
6aa8b732 AK |
3046 | |
3047 | static __init int kvm_init(void) | |
3048 | { | |
37e29d90 AK |
3049 | int r; |
3050 | ||
b5a33a75 AK |
3051 | r = kvm_mmu_module_init(); |
3052 | if (r) | |
3053 | goto out4; | |
3054 | ||
6aa8b732 AK |
3055 | kvm_init_debug(); |
3056 | ||
043405e1 | 3057 | kvm_arch_init(); |
bf591b24 | 3058 | |
cea7bb21 | 3059 | bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO); |
d77c26fc MD |
3060 | |
3061 | if (bad_page == NULL) { | |
6aa8b732 AK |
3062 | r = -ENOMEM; |
3063 | goto out; | |
3064 | } | |
3065 | ||
58e690e6 | 3066 | return 0; |
6aa8b732 AK |
3067 | |
3068 | out: | |
3069 | kvm_exit_debug(); | |
b5a33a75 AK |
3070 | kvm_mmu_module_exit(); |
3071 | out4: | |
6aa8b732 AK |
3072 | return r; |
3073 | } | |
3074 | ||
3075 | static __exit void kvm_exit(void) | |
3076 | { | |
3077 | kvm_exit_debug(); | |
cea7bb21 | 3078 | __free_page(bad_page); |
b5a33a75 | 3079 | kvm_mmu_module_exit(); |
6aa8b732 AK |
3080 | } |
3081 | ||
3082 | module_init(kvm_init) | |
3083 | module_exit(kvm_exit) |