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" | |
e495606d AK |
19 | #include "x86_emulate.h" |
20 | #include "segment_descriptor.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> |
6aa8b732 | 41 | |
e495606d AK |
42 | #include <asm/processor.h> |
43 | #include <asm/msr.h> | |
44 | #include <asm/io.h> | |
45 | #include <asm/uaccess.h> | |
46 | #include <asm/desc.h> | |
6aa8b732 AK |
47 | |
48 | MODULE_AUTHOR("Qumranet"); | |
49 | MODULE_LICENSE("GPL"); | |
50 | ||
133de902 AK |
51 | static DEFINE_SPINLOCK(kvm_lock); |
52 | static LIST_HEAD(vm_list); | |
53 | ||
1b6c0168 AK |
54 | static cpumask_t cpus_hardware_enabled; |
55 | ||
6aa8b732 | 56 | struct kvm_arch_ops *kvm_arch_ops; |
c16f862d RR |
57 | struct kmem_cache *kvm_vcpu_cache; |
58 | EXPORT_SYMBOL_GPL(kvm_vcpu_cache); | |
1165f5fe | 59 | |
15ad7146 AK |
60 | static __read_mostly struct preempt_ops kvm_preempt_ops; |
61 | ||
1165f5fe | 62 | #define STAT_OFFSET(x) offsetof(struct kvm_vcpu, stat.x) |
6aa8b732 AK |
63 | |
64 | static struct kvm_stats_debugfs_item { | |
65 | const char *name; | |
1165f5fe | 66 | int offset; |
6aa8b732 AK |
67 | struct dentry *dentry; |
68 | } debugfs_entries[] = { | |
1165f5fe AK |
69 | { "pf_fixed", STAT_OFFSET(pf_fixed) }, |
70 | { "pf_guest", STAT_OFFSET(pf_guest) }, | |
71 | { "tlb_flush", STAT_OFFSET(tlb_flush) }, | |
72 | { "invlpg", STAT_OFFSET(invlpg) }, | |
73 | { "exits", STAT_OFFSET(exits) }, | |
74 | { "io_exits", STAT_OFFSET(io_exits) }, | |
75 | { "mmio_exits", STAT_OFFSET(mmio_exits) }, | |
76 | { "signal_exits", STAT_OFFSET(signal_exits) }, | |
77 | { "irq_window", STAT_OFFSET(irq_window_exits) }, | |
78 | { "halt_exits", STAT_OFFSET(halt_exits) }, | |
b6958ce4 | 79 | { "halt_wakeup", STAT_OFFSET(halt_wakeup) }, |
1165f5fe AK |
80 | { "request_irq", STAT_OFFSET(request_irq_exits) }, |
81 | { "irq_exits", STAT_OFFSET(irq_exits) }, | |
e6adf283 | 82 | { "light_exits", STAT_OFFSET(light_exits) }, |
2cc51560 | 83 | { "efer_reload", STAT_OFFSET(efer_reload) }, |
1165f5fe | 84 | { NULL } |
6aa8b732 AK |
85 | }; |
86 | ||
87 | static struct dentry *debugfs_dir; | |
88 | ||
89 | #define MAX_IO_MSRS 256 | |
90 | ||
707d92fa RR |
91 | #define CR0_RESERVED_BITS \ |
92 | (~(unsigned long)(X86_CR0_PE | X86_CR0_MP | X86_CR0_EM | X86_CR0_TS \ | |
93 | | X86_CR0_ET | X86_CR0_NE | X86_CR0_WP | X86_CR0_AM \ | |
94 | | X86_CR0_NW | X86_CR0_CD | X86_CR0_PG)) | |
66aee91a RR |
95 | #define CR4_RESERVED_BITS \ |
96 | (~(unsigned long)(X86_CR4_VME | X86_CR4_PVI | X86_CR4_TSD | X86_CR4_DE\ | |
97 | | X86_CR4_PSE | X86_CR4_PAE | X86_CR4_MCE \ | |
98 | | X86_CR4_PGE | X86_CR4_PCE | X86_CR4_OSFXSR \ | |
99 | | X86_CR4_OSXMMEXCPT | X86_CR4_VMXE)) | |
100 | ||
7075bc81 | 101 | #define CR8_RESERVED_BITS (~(unsigned long)X86_CR8_TPR) |
6aa8b732 AK |
102 | #define EFER_RESERVED_BITS 0xfffffffffffff2fe |
103 | ||
05b3e0c2 | 104 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
105 | // LDT or TSS descriptor in the GDT. 16 bytes. |
106 | struct segment_descriptor_64 { | |
107 | struct segment_descriptor s; | |
108 | u32 base_higher; | |
109 | u32 pad_zero; | |
110 | }; | |
111 | ||
112 | #endif | |
113 | ||
bccf2150 AK |
114 | static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl, |
115 | unsigned long arg); | |
116 | ||
6aa8b732 AK |
117 | unsigned long segment_base(u16 selector) |
118 | { | |
119 | struct descriptor_table gdt; | |
120 | struct segment_descriptor *d; | |
121 | unsigned long table_base; | |
122 | typedef unsigned long ul; | |
123 | unsigned long v; | |
124 | ||
125 | if (selector == 0) | |
126 | return 0; | |
127 | ||
128 | asm ("sgdt %0" : "=m"(gdt)); | |
129 | table_base = gdt.base; | |
130 | ||
131 | if (selector & 4) { /* from ldt */ | |
132 | u16 ldt_selector; | |
133 | ||
134 | asm ("sldt %0" : "=g"(ldt_selector)); | |
135 | table_base = segment_base(ldt_selector); | |
136 | } | |
137 | d = (struct segment_descriptor *)(table_base + (selector & ~7)); | |
138 | v = d->base_low | ((ul)d->base_mid << 16) | ((ul)d->base_high << 24); | |
05b3e0c2 | 139 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
140 | if (d->system == 0 |
141 | && (d->type == 2 || d->type == 9 || d->type == 11)) | |
142 | v |= ((ul)((struct segment_descriptor_64 *)d)->base_higher) << 32; | |
143 | #endif | |
144 | return v; | |
145 | } | |
146 | EXPORT_SYMBOL_GPL(segment_base); | |
147 | ||
5aacf0ca JM |
148 | static inline int valid_vcpu(int n) |
149 | { | |
150 | return likely(n >= 0 && n < KVM_MAX_VCPUS); | |
151 | } | |
152 | ||
7702fd1f AK |
153 | void kvm_load_guest_fpu(struct kvm_vcpu *vcpu) |
154 | { | |
155 | if (!vcpu->fpu_active || vcpu->guest_fpu_loaded) | |
156 | return; | |
157 | ||
158 | vcpu->guest_fpu_loaded = 1; | |
b114b080 RR |
159 | fx_save(&vcpu->host_fx_image); |
160 | fx_restore(&vcpu->guest_fx_image); | |
7702fd1f AK |
161 | } |
162 | EXPORT_SYMBOL_GPL(kvm_load_guest_fpu); | |
163 | ||
164 | void kvm_put_guest_fpu(struct kvm_vcpu *vcpu) | |
165 | { | |
166 | if (!vcpu->guest_fpu_loaded) | |
167 | return; | |
168 | ||
169 | vcpu->guest_fpu_loaded = 0; | |
b114b080 RR |
170 | fx_save(&vcpu->guest_fx_image); |
171 | fx_restore(&vcpu->host_fx_image); | |
7702fd1f AK |
172 | } |
173 | EXPORT_SYMBOL_GPL(kvm_put_guest_fpu); | |
174 | ||
bccf2150 AK |
175 | /* |
176 | * Switches to specified vcpu, until a matching vcpu_put() | |
177 | */ | |
178 | static void vcpu_load(struct kvm_vcpu *vcpu) | |
6aa8b732 | 179 | { |
15ad7146 AK |
180 | int cpu; |
181 | ||
bccf2150 | 182 | mutex_lock(&vcpu->mutex); |
15ad7146 AK |
183 | cpu = get_cpu(); |
184 | preempt_notifier_register(&vcpu->preempt_notifier); | |
185 | kvm_arch_ops->vcpu_load(vcpu, cpu); | |
186 | put_cpu(); | |
6aa8b732 AK |
187 | } |
188 | ||
6aa8b732 AK |
189 | static void vcpu_put(struct kvm_vcpu *vcpu) |
190 | { | |
15ad7146 | 191 | preempt_disable(); |
6aa8b732 | 192 | kvm_arch_ops->vcpu_put(vcpu); |
15ad7146 AK |
193 | preempt_notifier_unregister(&vcpu->preempt_notifier); |
194 | preempt_enable(); | |
6aa8b732 AK |
195 | mutex_unlock(&vcpu->mutex); |
196 | } | |
197 | ||
d9e368d6 AK |
198 | static void ack_flush(void *_completed) |
199 | { | |
200 | atomic_t *completed = _completed; | |
201 | ||
202 | atomic_inc(completed); | |
203 | } | |
204 | ||
205 | void kvm_flush_remote_tlbs(struct kvm *kvm) | |
206 | { | |
207 | int i, cpu, needed; | |
208 | cpumask_t cpus; | |
209 | struct kvm_vcpu *vcpu; | |
210 | atomic_t completed; | |
211 | ||
212 | atomic_set(&completed, 0); | |
213 | cpus_clear(cpus); | |
214 | needed = 0; | |
fb3f0f51 RR |
215 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
216 | vcpu = kvm->vcpus[i]; | |
217 | if (!vcpu) | |
218 | continue; | |
d9e368d6 AK |
219 | if (test_and_set_bit(KVM_TLB_FLUSH, &vcpu->requests)) |
220 | continue; | |
221 | cpu = vcpu->cpu; | |
222 | if (cpu != -1 && cpu != raw_smp_processor_id()) | |
223 | if (!cpu_isset(cpu, cpus)) { | |
224 | cpu_set(cpu, cpus); | |
225 | ++needed; | |
226 | } | |
227 | } | |
228 | ||
229 | /* | |
230 | * We really want smp_call_function_mask() here. But that's not | |
231 | * available, so ipi all cpus in parallel and wait for them | |
232 | * to complete. | |
233 | */ | |
234 | for (cpu = first_cpu(cpus); cpu != NR_CPUS; cpu = next_cpu(cpu, cpus)) | |
235 | smp_call_function_single(cpu, ack_flush, &completed, 1, 0); | |
236 | while (atomic_read(&completed) != needed) { | |
237 | cpu_relax(); | |
238 | barrier(); | |
239 | } | |
240 | } | |
241 | ||
fb3f0f51 RR |
242 | int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id) |
243 | { | |
244 | struct page *page; | |
245 | int r; | |
246 | ||
247 | mutex_init(&vcpu->mutex); | |
248 | vcpu->cpu = -1; | |
249 | vcpu->mmu.root_hpa = INVALID_PAGE; | |
250 | vcpu->kvm = kvm; | |
251 | vcpu->vcpu_id = id; | |
b6958ce4 | 252 | init_waitqueue_head(&vcpu->wq); |
fb3f0f51 RR |
253 | |
254 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
255 | if (!page) { | |
256 | r = -ENOMEM; | |
257 | goto fail; | |
258 | } | |
259 | vcpu->run = page_address(page); | |
260 | ||
261 | page = alloc_page(GFP_KERNEL | __GFP_ZERO); | |
262 | if (!page) { | |
263 | r = -ENOMEM; | |
264 | goto fail_free_run; | |
265 | } | |
266 | vcpu->pio_data = page_address(page); | |
267 | ||
fb3f0f51 RR |
268 | r = kvm_mmu_create(vcpu); |
269 | if (r < 0) | |
270 | goto fail_free_pio_data; | |
271 | ||
272 | return 0; | |
273 | ||
274 | fail_free_pio_data: | |
275 | free_page((unsigned long)vcpu->pio_data); | |
276 | fail_free_run: | |
277 | free_page((unsigned long)vcpu->run); | |
278 | fail: | |
279 | return -ENOMEM; | |
280 | } | |
281 | EXPORT_SYMBOL_GPL(kvm_vcpu_init); | |
282 | ||
283 | void kvm_vcpu_uninit(struct kvm_vcpu *vcpu) | |
284 | { | |
285 | kvm_mmu_destroy(vcpu); | |
1b9778da ED |
286 | if (vcpu->apic) |
287 | hrtimer_cancel(&vcpu->apic->timer.dev); | |
97222cc8 | 288 | kvm_free_apic(vcpu->apic); |
fb3f0f51 RR |
289 | free_page((unsigned long)vcpu->pio_data); |
290 | free_page((unsigned long)vcpu->run); | |
291 | } | |
292 | EXPORT_SYMBOL_GPL(kvm_vcpu_uninit); | |
293 | ||
f17abe9a | 294 | static struct kvm *kvm_create_vm(void) |
6aa8b732 AK |
295 | { |
296 | struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL); | |
6aa8b732 AK |
297 | |
298 | if (!kvm) | |
f17abe9a | 299 | return ERR_PTR(-ENOMEM); |
6aa8b732 | 300 | |
74906345 | 301 | kvm_io_bus_init(&kvm->pio_bus); |
11ec2804 | 302 | mutex_init(&kvm->lock); |
6aa8b732 | 303 | INIT_LIST_HEAD(&kvm->active_mmu_pages); |
2eeb2e94 | 304 | kvm_io_bus_init(&kvm->mmio_bus); |
5e58cfe4 RR |
305 | spin_lock(&kvm_lock); |
306 | list_add(&kvm->vm_list, &vm_list); | |
307 | spin_unlock(&kvm_lock); | |
f17abe9a AK |
308 | return kvm; |
309 | } | |
310 | ||
6aa8b732 AK |
311 | /* |
312 | * Free any memory in @free but not in @dont. | |
313 | */ | |
314 | static void kvm_free_physmem_slot(struct kvm_memory_slot *free, | |
315 | struct kvm_memory_slot *dont) | |
316 | { | |
317 | int i; | |
318 | ||
319 | if (!dont || free->phys_mem != dont->phys_mem) | |
320 | if (free->phys_mem) { | |
321 | for (i = 0; i < free->npages; ++i) | |
55a54f79 AK |
322 | if (free->phys_mem[i]) |
323 | __free_page(free->phys_mem[i]); | |
6aa8b732 AK |
324 | vfree(free->phys_mem); |
325 | } | |
326 | ||
327 | if (!dont || free->dirty_bitmap != dont->dirty_bitmap) | |
328 | vfree(free->dirty_bitmap); | |
329 | ||
8b6d44c7 | 330 | free->phys_mem = NULL; |
6aa8b732 | 331 | free->npages = 0; |
8b6d44c7 | 332 | free->dirty_bitmap = NULL; |
6aa8b732 AK |
333 | } |
334 | ||
335 | static void kvm_free_physmem(struct kvm *kvm) | |
336 | { | |
337 | int i; | |
338 | ||
339 | for (i = 0; i < kvm->nmemslots; ++i) | |
8b6d44c7 | 340 | kvm_free_physmem_slot(&kvm->memslots[i], NULL); |
6aa8b732 AK |
341 | } |
342 | ||
039576c0 AK |
343 | static void free_pio_guest_pages(struct kvm_vcpu *vcpu) |
344 | { | |
345 | int i; | |
346 | ||
3077c451 | 347 | for (i = 0; i < ARRAY_SIZE(vcpu->pio.guest_pages); ++i) |
039576c0 AK |
348 | if (vcpu->pio.guest_pages[i]) { |
349 | __free_page(vcpu->pio.guest_pages[i]); | |
350 | vcpu->pio.guest_pages[i] = NULL; | |
351 | } | |
352 | } | |
353 | ||
7b53aa56 AK |
354 | static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu) |
355 | { | |
7b53aa56 AK |
356 | vcpu_load(vcpu); |
357 | kvm_mmu_unload(vcpu); | |
358 | vcpu_put(vcpu); | |
359 | } | |
360 | ||
6aa8b732 AK |
361 | static void kvm_free_vcpus(struct kvm *kvm) |
362 | { | |
363 | unsigned int i; | |
364 | ||
7b53aa56 AK |
365 | /* |
366 | * Unpin any mmu pages first. | |
367 | */ | |
368 | for (i = 0; i < KVM_MAX_VCPUS; ++i) | |
fb3f0f51 RR |
369 | if (kvm->vcpus[i]) |
370 | kvm_unload_vcpu_mmu(kvm->vcpus[i]); | |
371 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
372 | if (kvm->vcpus[i]) { | |
373 | kvm_arch_ops->vcpu_free(kvm->vcpus[i]); | |
374 | kvm->vcpus[i] = NULL; | |
375 | } | |
376 | } | |
377 | ||
6aa8b732 AK |
378 | } |
379 | ||
f17abe9a AK |
380 | static void kvm_destroy_vm(struct kvm *kvm) |
381 | { | |
133de902 AK |
382 | spin_lock(&kvm_lock); |
383 | list_del(&kvm->vm_list); | |
384 | spin_unlock(&kvm_lock); | |
74906345 | 385 | kvm_io_bus_destroy(&kvm->pio_bus); |
2eeb2e94 | 386 | kvm_io_bus_destroy(&kvm->mmio_bus); |
85f455f7 | 387 | kfree(kvm->vpic); |
1fd4f2a5 | 388 | kfree(kvm->vioapic); |
6aa8b732 AK |
389 | kvm_free_vcpus(kvm); |
390 | kvm_free_physmem(kvm); | |
391 | kfree(kvm); | |
f17abe9a AK |
392 | } |
393 | ||
394 | static int kvm_vm_release(struct inode *inode, struct file *filp) | |
395 | { | |
396 | struct kvm *kvm = filp->private_data; | |
397 | ||
398 | kvm_destroy_vm(kvm); | |
6aa8b732 AK |
399 | return 0; |
400 | } | |
401 | ||
402 | static void inject_gp(struct kvm_vcpu *vcpu) | |
403 | { | |
404 | kvm_arch_ops->inject_gp(vcpu, 0); | |
405 | } | |
406 | ||
1342d353 AK |
407 | /* |
408 | * Load the pae pdptrs. Return true is they are all valid. | |
409 | */ | |
410 | static int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3) | |
6aa8b732 AK |
411 | { |
412 | gfn_t pdpt_gfn = cr3 >> PAGE_SHIFT; | |
1342d353 | 413 | unsigned offset = ((cr3 & (PAGE_SIZE-1)) >> 5) << 2; |
6aa8b732 | 414 | int i; |
6aa8b732 | 415 | u64 *pdpt; |
1342d353 | 416 | int ret; |
954bbbc2 | 417 | struct page *page; |
c820c2aa | 418 | u64 pdpte[ARRAY_SIZE(vcpu->pdptrs)]; |
6aa8b732 | 419 | |
11ec2804 | 420 | mutex_lock(&vcpu->kvm->lock); |
954bbbc2 | 421 | page = gfn_to_page(vcpu->kvm, pdpt_gfn); |
c820c2aa RR |
422 | if (!page) { |
423 | ret = 0; | |
424 | goto out; | |
425 | } | |
426 | ||
954bbbc2 | 427 | pdpt = kmap_atomic(page, KM_USER0); |
c820c2aa RR |
428 | memcpy(pdpte, pdpt+offset, sizeof(pdpte)); |
429 | kunmap_atomic(pdpt, KM_USER0); | |
6aa8b732 | 430 | |
c820c2aa RR |
431 | for (i = 0; i < ARRAY_SIZE(pdpte); ++i) { |
432 | if ((pdpte[i] & 1) && (pdpte[i] & 0xfffffff0000001e6ull)) { | |
1342d353 AK |
433 | ret = 0; |
434 | goto out; | |
435 | } | |
6aa8b732 | 436 | } |
c820c2aa | 437 | ret = 1; |
6aa8b732 | 438 | |
c820c2aa | 439 | memcpy(vcpu->pdptrs, pdpte, sizeof(vcpu->pdptrs)); |
1342d353 | 440 | out: |
11ec2804 | 441 | mutex_unlock(&vcpu->kvm->lock); |
6aa8b732 | 442 | |
1342d353 | 443 | return ret; |
6aa8b732 AK |
444 | } |
445 | ||
446 | void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0) | |
447 | { | |
707d92fa | 448 | if (cr0 & CR0_RESERVED_BITS) { |
6aa8b732 AK |
449 | printk(KERN_DEBUG "set_cr0: 0x%lx #GP, reserved bits 0x%lx\n", |
450 | cr0, vcpu->cr0); | |
451 | inject_gp(vcpu); | |
452 | return; | |
453 | } | |
454 | ||
707d92fa | 455 | if ((cr0 & X86_CR0_NW) && !(cr0 & X86_CR0_CD)) { |
6aa8b732 AK |
456 | printk(KERN_DEBUG "set_cr0: #GP, CD == 0 && NW == 1\n"); |
457 | inject_gp(vcpu); | |
458 | return; | |
459 | } | |
460 | ||
707d92fa | 461 | if ((cr0 & X86_CR0_PG) && !(cr0 & X86_CR0_PE)) { |
6aa8b732 AK |
462 | printk(KERN_DEBUG "set_cr0: #GP, set PG flag " |
463 | "and a clear PE flag\n"); | |
464 | inject_gp(vcpu); | |
465 | return; | |
466 | } | |
467 | ||
707d92fa | 468 | if (!is_paging(vcpu) && (cr0 & X86_CR0_PG)) { |
05b3e0c2 | 469 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
470 | if ((vcpu->shadow_efer & EFER_LME)) { |
471 | int cs_db, cs_l; | |
472 | ||
473 | if (!is_pae(vcpu)) { | |
474 | printk(KERN_DEBUG "set_cr0: #GP, start paging " | |
475 | "in long mode while PAE is disabled\n"); | |
476 | inject_gp(vcpu); | |
477 | return; | |
478 | } | |
479 | kvm_arch_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l); | |
480 | if (cs_l) { | |
481 | printk(KERN_DEBUG "set_cr0: #GP, start paging " | |
482 | "in long mode while CS.L == 1\n"); | |
483 | inject_gp(vcpu); | |
484 | return; | |
485 | ||
486 | } | |
487 | } else | |
488 | #endif | |
1342d353 | 489 | if (is_pae(vcpu) && !load_pdptrs(vcpu, vcpu->cr3)) { |
6aa8b732 AK |
490 | printk(KERN_DEBUG "set_cr0: #GP, pdptrs " |
491 | "reserved bits\n"); | |
492 | inject_gp(vcpu); | |
493 | return; | |
494 | } | |
495 | ||
496 | } | |
497 | ||
498 | kvm_arch_ops->set_cr0(vcpu, cr0); | |
499 | vcpu->cr0 = cr0; | |
500 | ||
11ec2804 | 501 | mutex_lock(&vcpu->kvm->lock); |
6aa8b732 | 502 | kvm_mmu_reset_context(vcpu); |
11ec2804 | 503 | mutex_unlock(&vcpu->kvm->lock); |
6aa8b732 AK |
504 | return; |
505 | } | |
506 | EXPORT_SYMBOL_GPL(set_cr0); | |
507 | ||
508 | void lmsw(struct kvm_vcpu *vcpu, unsigned long msw) | |
509 | { | |
510 | set_cr0(vcpu, (vcpu->cr0 & ~0x0ful) | (msw & 0x0f)); | |
511 | } | |
512 | EXPORT_SYMBOL_GPL(lmsw); | |
513 | ||
514 | void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4) | |
515 | { | |
66aee91a | 516 | if (cr4 & CR4_RESERVED_BITS) { |
6aa8b732 AK |
517 | printk(KERN_DEBUG "set_cr4: #GP, reserved bits\n"); |
518 | inject_gp(vcpu); | |
519 | return; | |
520 | } | |
521 | ||
a9058ecd | 522 | if (is_long_mode(vcpu)) { |
66aee91a | 523 | if (!(cr4 & X86_CR4_PAE)) { |
6aa8b732 AK |
524 | printk(KERN_DEBUG "set_cr4: #GP, clearing PAE while " |
525 | "in long mode\n"); | |
526 | inject_gp(vcpu); | |
527 | return; | |
528 | } | |
66aee91a | 529 | } else if (is_paging(vcpu) && !is_pae(vcpu) && (cr4 & X86_CR4_PAE) |
1342d353 | 530 | && !load_pdptrs(vcpu, vcpu->cr3)) { |
6aa8b732 AK |
531 | printk(KERN_DEBUG "set_cr4: #GP, pdptrs reserved bits\n"); |
532 | inject_gp(vcpu); | |
310bc76c | 533 | return; |
6aa8b732 AK |
534 | } |
535 | ||
66aee91a | 536 | if (cr4 & X86_CR4_VMXE) { |
6aa8b732 AK |
537 | printk(KERN_DEBUG "set_cr4: #GP, setting VMXE\n"); |
538 | inject_gp(vcpu); | |
539 | return; | |
540 | } | |
541 | kvm_arch_ops->set_cr4(vcpu, cr4); | |
11ec2804 | 542 | mutex_lock(&vcpu->kvm->lock); |
6aa8b732 | 543 | kvm_mmu_reset_context(vcpu); |
11ec2804 | 544 | mutex_unlock(&vcpu->kvm->lock); |
6aa8b732 AK |
545 | } |
546 | EXPORT_SYMBOL_GPL(set_cr4); | |
547 | ||
548 | void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3) | |
549 | { | |
a9058ecd | 550 | if (is_long_mode(vcpu)) { |
f802a307 | 551 | if (cr3 & CR3_L_MODE_RESERVED_BITS) { |
6aa8b732 AK |
552 | printk(KERN_DEBUG "set_cr3: #GP, reserved bits\n"); |
553 | inject_gp(vcpu); | |
554 | return; | |
555 | } | |
556 | } else { | |
f802a307 RR |
557 | if (is_pae(vcpu)) { |
558 | if (cr3 & CR3_PAE_RESERVED_BITS) { | |
559 | printk(KERN_DEBUG | |
560 | "set_cr3: #GP, reserved bits\n"); | |
561 | inject_gp(vcpu); | |
562 | return; | |
563 | } | |
564 | if (is_paging(vcpu) && !load_pdptrs(vcpu, cr3)) { | |
565 | printk(KERN_DEBUG "set_cr3: #GP, pdptrs " | |
566 | "reserved bits\n"); | |
567 | inject_gp(vcpu); | |
568 | return; | |
569 | } | |
570 | } else { | |
571 | if (cr3 & CR3_NONPAE_RESERVED_BITS) { | |
572 | printk(KERN_DEBUG | |
573 | "set_cr3: #GP, reserved bits\n"); | |
574 | inject_gp(vcpu); | |
575 | return; | |
576 | } | |
6aa8b732 AK |
577 | } |
578 | } | |
579 | ||
11ec2804 | 580 | mutex_lock(&vcpu->kvm->lock); |
d21225ee IM |
581 | /* |
582 | * Does the new cr3 value map to physical memory? (Note, we | |
583 | * catch an invalid cr3 even in real-mode, because it would | |
584 | * cause trouble later on when we turn on paging anyway.) | |
585 | * | |
586 | * A real CPU would silently accept an invalid cr3 and would | |
587 | * attempt to use it - with largely undefined (and often hard | |
588 | * to debug) behavior on the guest side. | |
589 | */ | |
590 | if (unlikely(!gfn_to_memslot(vcpu->kvm, cr3 >> PAGE_SHIFT))) | |
591 | inject_gp(vcpu); | |
fb764416 RR |
592 | else { |
593 | vcpu->cr3 = cr3; | |
d21225ee | 594 | vcpu->mmu.new_cr3(vcpu); |
fb764416 | 595 | } |
11ec2804 | 596 | mutex_unlock(&vcpu->kvm->lock); |
6aa8b732 AK |
597 | } |
598 | EXPORT_SYMBOL_GPL(set_cr3); | |
599 | ||
600 | void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8) | |
601 | { | |
7075bc81 | 602 | if (cr8 & CR8_RESERVED_BITS) { |
6aa8b732 AK |
603 | printk(KERN_DEBUG "set_cr8: #GP, reserved bits 0x%lx\n", cr8); |
604 | inject_gp(vcpu); | |
605 | return; | |
606 | } | |
97222cc8 ED |
607 | if (irqchip_in_kernel(vcpu->kvm)) |
608 | kvm_lapic_set_tpr(vcpu, cr8); | |
609 | else | |
610 | vcpu->cr8 = cr8; | |
6aa8b732 AK |
611 | } |
612 | EXPORT_SYMBOL_GPL(set_cr8); | |
613 | ||
7017fc3d ED |
614 | unsigned long get_cr8(struct kvm_vcpu *vcpu) |
615 | { | |
97222cc8 ED |
616 | if (irqchip_in_kernel(vcpu->kvm)) |
617 | return kvm_lapic_get_cr8(vcpu); | |
618 | else | |
619 | return vcpu->cr8; | |
7017fc3d ED |
620 | } |
621 | EXPORT_SYMBOL_GPL(get_cr8); | |
622 | ||
623 | u64 kvm_get_apic_base(struct kvm_vcpu *vcpu) | |
624 | { | |
97222cc8 ED |
625 | if (irqchip_in_kernel(vcpu->kvm)) |
626 | return vcpu->apic_base; | |
627 | else | |
628 | return vcpu->apic_base; | |
7017fc3d ED |
629 | } |
630 | EXPORT_SYMBOL_GPL(kvm_get_apic_base); | |
631 | ||
632 | void kvm_set_apic_base(struct kvm_vcpu *vcpu, u64 data) | |
633 | { | |
97222cc8 ED |
634 | /* TODO: reserve bits check */ |
635 | if (irqchip_in_kernel(vcpu->kvm)) | |
636 | kvm_lapic_set_base(vcpu, data); | |
637 | else | |
638 | vcpu->apic_base = data; | |
7017fc3d ED |
639 | } |
640 | EXPORT_SYMBOL_GPL(kvm_set_apic_base); | |
641 | ||
6aa8b732 AK |
642 | void fx_init(struct kvm_vcpu *vcpu) |
643 | { | |
b114b080 | 644 | unsigned after_mxcsr_mask; |
6aa8b732 | 645 | |
9bd01506 RR |
646 | /* Initialize guest FPU by resetting ours and saving into guest's */ |
647 | preempt_disable(); | |
b114b080 | 648 | fx_save(&vcpu->host_fx_image); |
6aa8b732 | 649 | fpu_init(); |
b114b080 RR |
650 | fx_save(&vcpu->guest_fx_image); |
651 | fx_restore(&vcpu->host_fx_image); | |
9bd01506 | 652 | preempt_enable(); |
6aa8b732 | 653 | |
b114b080 RR |
654 | after_mxcsr_mask = offsetof(struct i387_fxsave_struct, st_space); |
655 | vcpu->guest_fx_image.mxcsr = 0x1f80; | |
656 | memset((void *)&vcpu->guest_fx_image + after_mxcsr_mask, | |
657 | 0, sizeof(struct i387_fxsave_struct) - after_mxcsr_mask); | |
6aa8b732 AK |
658 | } |
659 | EXPORT_SYMBOL_GPL(fx_init); | |
660 | ||
6aa8b732 AK |
661 | /* |
662 | * Allocate some memory and give it an address in the guest physical address | |
663 | * space. | |
664 | * | |
665 | * Discontiguous memory is allowed, mostly for framebuffers. | |
666 | */ | |
2c6f5df9 AK |
667 | static int kvm_vm_ioctl_set_memory_region(struct kvm *kvm, |
668 | struct kvm_memory_region *mem) | |
6aa8b732 AK |
669 | { |
670 | int r; | |
671 | gfn_t base_gfn; | |
672 | unsigned long npages; | |
673 | unsigned long i; | |
674 | struct kvm_memory_slot *memslot; | |
675 | struct kvm_memory_slot old, new; | |
676 | int memory_config_version; | |
677 | ||
678 | r = -EINVAL; | |
679 | /* General sanity checks */ | |
680 | if (mem->memory_size & (PAGE_SIZE - 1)) | |
681 | goto out; | |
682 | if (mem->guest_phys_addr & (PAGE_SIZE - 1)) | |
683 | goto out; | |
684 | if (mem->slot >= KVM_MEMORY_SLOTS) | |
685 | goto out; | |
686 | if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr) | |
687 | goto out; | |
688 | ||
689 | memslot = &kvm->memslots[mem->slot]; | |
690 | base_gfn = mem->guest_phys_addr >> PAGE_SHIFT; | |
691 | npages = mem->memory_size >> PAGE_SHIFT; | |
692 | ||
693 | if (!npages) | |
694 | mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES; | |
695 | ||
696 | raced: | |
11ec2804 | 697 | mutex_lock(&kvm->lock); |
6aa8b732 AK |
698 | |
699 | memory_config_version = kvm->memory_config_version; | |
700 | new = old = *memslot; | |
701 | ||
702 | new.base_gfn = base_gfn; | |
703 | new.npages = npages; | |
704 | new.flags = mem->flags; | |
705 | ||
706 | /* Disallow changing a memory slot's size. */ | |
707 | r = -EINVAL; | |
708 | if (npages && old.npages && npages != old.npages) | |
709 | goto out_unlock; | |
710 | ||
711 | /* Check for overlaps */ | |
712 | r = -EEXIST; | |
713 | for (i = 0; i < KVM_MEMORY_SLOTS; ++i) { | |
714 | struct kvm_memory_slot *s = &kvm->memslots[i]; | |
715 | ||
716 | if (s == memslot) | |
717 | continue; | |
718 | if (!((base_gfn + npages <= s->base_gfn) || | |
719 | (base_gfn >= s->base_gfn + s->npages))) | |
720 | goto out_unlock; | |
721 | } | |
722 | /* | |
723 | * Do memory allocations outside lock. memory_config_version will | |
724 | * detect any races. | |
725 | */ | |
11ec2804 | 726 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
727 | |
728 | /* Deallocate if slot is being removed */ | |
729 | if (!npages) | |
8b6d44c7 | 730 | new.phys_mem = NULL; |
6aa8b732 AK |
731 | |
732 | /* Free page dirty bitmap if unneeded */ | |
733 | if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES)) | |
8b6d44c7 | 734 | new.dirty_bitmap = NULL; |
6aa8b732 AK |
735 | |
736 | r = -ENOMEM; | |
737 | ||
738 | /* Allocate if a slot is being created */ | |
739 | if (npages && !new.phys_mem) { | |
740 | new.phys_mem = vmalloc(npages * sizeof(struct page *)); | |
741 | ||
742 | if (!new.phys_mem) | |
743 | goto out_free; | |
744 | ||
745 | memset(new.phys_mem, 0, npages * sizeof(struct page *)); | |
746 | for (i = 0; i < npages; ++i) { | |
747 | new.phys_mem[i] = alloc_page(GFP_HIGHUSER | |
748 | | __GFP_ZERO); | |
749 | if (!new.phys_mem[i]) | |
750 | goto out_free; | |
5972e953 | 751 | set_page_private(new.phys_mem[i],0); |
6aa8b732 AK |
752 | } |
753 | } | |
754 | ||
755 | /* Allocate page dirty bitmap if needed */ | |
756 | if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) { | |
757 | unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8; | |
758 | ||
759 | new.dirty_bitmap = vmalloc(dirty_bytes); | |
760 | if (!new.dirty_bitmap) | |
761 | goto out_free; | |
762 | memset(new.dirty_bitmap, 0, dirty_bytes); | |
763 | } | |
764 | ||
11ec2804 | 765 | mutex_lock(&kvm->lock); |
6aa8b732 AK |
766 | |
767 | if (memory_config_version != kvm->memory_config_version) { | |
11ec2804 | 768 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
769 | kvm_free_physmem_slot(&new, &old); |
770 | goto raced; | |
771 | } | |
772 | ||
773 | r = -EAGAIN; | |
774 | if (kvm->busy) | |
775 | goto out_unlock; | |
776 | ||
777 | if (mem->slot >= kvm->nmemslots) | |
778 | kvm->nmemslots = mem->slot + 1; | |
779 | ||
780 | *memslot = new; | |
781 | ++kvm->memory_config_version; | |
782 | ||
90cb0529 AK |
783 | kvm_mmu_slot_remove_write_access(kvm, mem->slot); |
784 | kvm_flush_remote_tlbs(kvm); | |
6aa8b732 | 785 | |
11ec2804 | 786 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
787 | |
788 | kvm_free_physmem_slot(&old, &new); | |
789 | return 0; | |
790 | ||
791 | out_unlock: | |
11ec2804 | 792 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
793 | out_free: |
794 | kvm_free_physmem_slot(&new, &old); | |
795 | out: | |
796 | return r; | |
797 | } | |
798 | ||
799 | /* | |
800 | * Get (and clear) the dirty memory log for a memory slot. | |
801 | */ | |
2c6f5df9 AK |
802 | static int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm, |
803 | struct kvm_dirty_log *log) | |
6aa8b732 AK |
804 | { |
805 | struct kvm_memory_slot *memslot; | |
806 | int r, i; | |
807 | int n; | |
808 | unsigned long any = 0; | |
809 | ||
11ec2804 | 810 | mutex_lock(&kvm->lock); |
6aa8b732 AK |
811 | |
812 | /* | |
813 | * Prevent changes to guest memory configuration even while the lock | |
814 | * is not taken. | |
815 | */ | |
816 | ++kvm->busy; | |
11ec2804 | 817 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
818 | r = -EINVAL; |
819 | if (log->slot >= KVM_MEMORY_SLOTS) | |
820 | goto out; | |
821 | ||
822 | memslot = &kvm->memslots[log->slot]; | |
823 | r = -ENOENT; | |
824 | if (!memslot->dirty_bitmap) | |
825 | goto out; | |
826 | ||
cd1a4a98 | 827 | n = ALIGN(memslot->npages, BITS_PER_LONG) / 8; |
6aa8b732 | 828 | |
cd1a4a98 | 829 | for (i = 0; !any && i < n/sizeof(long); ++i) |
6aa8b732 AK |
830 | any = memslot->dirty_bitmap[i]; |
831 | ||
832 | r = -EFAULT; | |
833 | if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n)) | |
834 | goto out; | |
835 | ||
39214915 RR |
836 | /* If nothing is dirty, don't bother messing with page tables. */ |
837 | if (any) { | |
838 | mutex_lock(&kvm->lock); | |
839 | kvm_mmu_slot_remove_write_access(kvm, log->slot); | |
840 | kvm_flush_remote_tlbs(kvm); | |
841 | memset(memslot->dirty_bitmap, 0, n); | |
842 | mutex_unlock(&kvm->lock); | |
843 | } | |
6aa8b732 AK |
844 | |
845 | r = 0; | |
846 | ||
847 | out: | |
11ec2804 | 848 | mutex_lock(&kvm->lock); |
6aa8b732 | 849 | --kvm->busy; |
11ec2804 | 850 | mutex_unlock(&kvm->lock); |
6aa8b732 AK |
851 | return r; |
852 | } | |
853 | ||
e8207547 AK |
854 | /* |
855 | * Set a new alias region. Aliases map a portion of physical memory into | |
856 | * another portion. This is useful for memory windows, for example the PC | |
857 | * VGA region. | |
858 | */ | |
859 | static int kvm_vm_ioctl_set_memory_alias(struct kvm *kvm, | |
860 | struct kvm_memory_alias *alias) | |
861 | { | |
862 | int r, n; | |
863 | struct kvm_mem_alias *p; | |
864 | ||
865 | r = -EINVAL; | |
866 | /* General sanity checks */ | |
867 | if (alias->memory_size & (PAGE_SIZE - 1)) | |
868 | goto out; | |
869 | if (alias->guest_phys_addr & (PAGE_SIZE - 1)) | |
870 | goto out; | |
871 | if (alias->slot >= KVM_ALIAS_SLOTS) | |
872 | goto out; | |
873 | if (alias->guest_phys_addr + alias->memory_size | |
874 | < alias->guest_phys_addr) | |
875 | goto out; | |
876 | if (alias->target_phys_addr + alias->memory_size | |
877 | < alias->target_phys_addr) | |
878 | goto out; | |
879 | ||
11ec2804 | 880 | mutex_lock(&kvm->lock); |
e8207547 AK |
881 | |
882 | p = &kvm->aliases[alias->slot]; | |
883 | p->base_gfn = alias->guest_phys_addr >> PAGE_SHIFT; | |
884 | p->npages = alias->memory_size >> PAGE_SHIFT; | |
885 | p->target_gfn = alias->target_phys_addr >> PAGE_SHIFT; | |
886 | ||
887 | for (n = KVM_ALIAS_SLOTS; n > 0; --n) | |
888 | if (kvm->aliases[n - 1].npages) | |
889 | break; | |
890 | kvm->naliases = n; | |
891 | ||
90cb0529 | 892 | kvm_mmu_zap_all(kvm); |
e8207547 | 893 | |
11ec2804 | 894 | mutex_unlock(&kvm->lock); |
e8207547 AK |
895 | |
896 | return 0; | |
897 | ||
898 | out: | |
899 | return r; | |
900 | } | |
901 | ||
6ceb9d79 HQ |
902 | static int kvm_vm_ioctl_get_irqchip(struct kvm *kvm, struct kvm_irqchip *chip) |
903 | { | |
904 | int r; | |
905 | ||
906 | r = 0; | |
907 | switch (chip->chip_id) { | |
908 | case KVM_IRQCHIP_PIC_MASTER: | |
909 | memcpy (&chip->chip.pic, | |
910 | &pic_irqchip(kvm)->pics[0], | |
911 | sizeof(struct kvm_pic_state)); | |
912 | break; | |
913 | case KVM_IRQCHIP_PIC_SLAVE: | |
914 | memcpy (&chip->chip.pic, | |
915 | &pic_irqchip(kvm)->pics[1], | |
916 | sizeof(struct kvm_pic_state)); | |
917 | break; | |
6bf9e962 HQ |
918 | case KVM_IRQCHIP_IOAPIC: |
919 | memcpy (&chip->chip.ioapic, | |
920 | ioapic_irqchip(kvm), | |
921 | sizeof(struct kvm_ioapic_state)); | |
922 | break; | |
6ceb9d79 HQ |
923 | default: |
924 | r = -EINVAL; | |
925 | break; | |
926 | } | |
927 | return r; | |
928 | } | |
929 | ||
930 | static int kvm_vm_ioctl_set_irqchip(struct kvm *kvm, struct kvm_irqchip *chip) | |
931 | { | |
932 | int r; | |
933 | ||
934 | r = 0; | |
935 | switch (chip->chip_id) { | |
936 | case KVM_IRQCHIP_PIC_MASTER: | |
937 | memcpy (&pic_irqchip(kvm)->pics[0], | |
938 | &chip->chip.pic, | |
939 | sizeof(struct kvm_pic_state)); | |
940 | break; | |
941 | case KVM_IRQCHIP_PIC_SLAVE: | |
942 | memcpy (&pic_irqchip(kvm)->pics[1], | |
943 | &chip->chip.pic, | |
944 | sizeof(struct kvm_pic_state)); | |
945 | break; | |
6bf9e962 HQ |
946 | case KVM_IRQCHIP_IOAPIC: |
947 | memcpy (ioapic_irqchip(kvm), | |
948 | &chip->chip.ioapic, | |
949 | sizeof(struct kvm_ioapic_state)); | |
950 | break; | |
6ceb9d79 HQ |
951 | default: |
952 | r = -EINVAL; | |
953 | break; | |
954 | } | |
955 | kvm_pic_update_irq(pic_irqchip(kvm)); | |
956 | return r; | |
957 | } | |
958 | ||
e8207547 AK |
959 | static gfn_t unalias_gfn(struct kvm *kvm, gfn_t gfn) |
960 | { | |
961 | int i; | |
962 | struct kvm_mem_alias *alias; | |
963 | ||
964 | for (i = 0; i < kvm->naliases; ++i) { | |
965 | alias = &kvm->aliases[i]; | |
966 | if (gfn >= alias->base_gfn | |
967 | && gfn < alias->base_gfn + alias->npages) | |
968 | return alias->target_gfn + gfn - alias->base_gfn; | |
969 | } | |
970 | return gfn; | |
971 | } | |
972 | ||
973 | static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
6aa8b732 AK |
974 | { |
975 | int i; | |
976 | ||
977 | for (i = 0; i < kvm->nmemslots; ++i) { | |
978 | struct kvm_memory_slot *memslot = &kvm->memslots[i]; | |
979 | ||
980 | if (gfn >= memslot->base_gfn | |
981 | && gfn < memslot->base_gfn + memslot->npages) | |
982 | return memslot; | |
983 | } | |
8b6d44c7 | 984 | return NULL; |
6aa8b732 | 985 | } |
e8207547 AK |
986 | |
987 | struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn) | |
988 | { | |
989 | gfn = unalias_gfn(kvm, gfn); | |
990 | return __gfn_to_memslot(kvm, gfn); | |
991 | } | |
6aa8b732 | 992 | |
954bbbc2 AK |
993 | struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn) |
994 | { | |
995 | struct kvm_memory_slot *slot; | |
996 | ||
e8207547 AK |
997 | gfn = unalias_gfn(kvm, gfn); |
998 | slot = __gfn_to_memslot(kvm, gfn); | |
954bbbc2 AK |
999 | if (!slot) |
1000 | return NULL; | |
1001 | return slot->phys_mem[gfn - slot->base_gfn]; | |
1002 | } | |
1003 | EXPORT_SYMBOL_GPL(gfn_to_page); | |
1004 | ||
7e9d619d | 1005 | /* WARNING: Does not work on aliased pages. */ |
6aa8b732 AK |
1006 | void mark_page_dirty(struct kvm *kvm, gfn_t gfn) |
1007 | { | |
31389947 | 1008 | struct kvm_memory_slot *memslot; |
6aa8b732 | 1009 | |
7e9d619d RR |
1010 | memslot = __gfn_to_memslot(kvm, gfn); |
1011 | if (memslot && memslot->dirty_bitmap) { | |
1012 | unsigned long rel_gfn = gfn - memslot->base_gfn; | |
6aa8b732 | 1013 | |
7e9d619d RR |
1014 | /* avoid RMW */ |
1015 | if (!test_bit(rel_gfn, memslot->dirty_bitmap)) | |
1016 | set_bit(rel_gfn, memslot->dirty_bitmap); | |
6aa8b732 AK |
1017 | } |
1018 | } | |
1019 | ||
e7d5d76c | 1020 | int emulator_read_std(unsigned long addr, |
4c690a1e | 1021 | void *val, |
6aa8b732 | 1022 | unsigned int bytes, |
cebff02b | 1023 | struct kvm_vcpu *vcpu) |
6aa8b732 | 1024 | { |
6aa8b732 AK |
1025 | void *data = val; |
1026 | ||
1027 | while (bytes) { | |
1028 | gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr); | |
1029 | unsigned offset = addr & (PAGE_SIZE-1); | |
1030 | unsigned tocopy = min(bytes, (unsigned)PAGE_SIZE - offset); | |
1031 | unsigned long pfn; | |
954bbbc2 AK |
1032 | struct page *page; |
1033 | void *page_virt; | |
6aa8b732 AK |
1034 | |
1035 | if (gpa == UNMAPPED_GVA) | |
1036 | return X86EMUL_PROPAGATE_FAULT; | |
1037 | pfn = gpa >> PAGE_SHIFT; | |
954bbbc2 AK |
1038 | page = gfn_to_page(vcpu->kvm, pfn); |
1039 | if (!page) | |
6aa8b732 | 1040 | return X86EMUL_UNHANDLEABLE; |
954bbbc2 | 1041 | page_virt = kmap_atomic(page, KM_USER0); |
6aa8b732 | 1042 | |
954bbbc2 | 1043 | memcpy(data, page_virt + offset, tocopy); |
6aa8b732 | 1044 | |
954bbbc2 | 1045 | kunmap_atomic(page_virt, KM_USER0); |
6aa8b732 AK |
1046 | |
1047 | bytes -= tocopy; | |
1048 | data += tocopy; | |
1049 | addr += tocopy; | |
1050 | } | |
1051 | ||
1052 | return X86EMUL_CONTINUE; | |
1053 | } | |
e7d5d76c | 1054 | EXPORT_SYMBOL_GPL(emulator_read_std); |
6aa8b732 AK |
1055 | |
1056 | static int emulator_write_std(unsigned long addr, | |
4c690a1e | 1057 | const void *val, |
6aa8b732 | 1058 | unsigned int bytes, |
cebff02b | 1059 | struct kvm_vcpu *vcpu) |
6aa8b732 | 1060 | { |
f0242478 | 1061 | pr_unimpl(vcpu, "emulator_write_std: addr %lx n %d\n", addr, bytes); |
6aa8b732 AK |
1062 | return X86EMUL_UNHANDLEABLE; |
1063 | } | |
1064 | ||
97222cc8 ED |
1065 | /* |
1066 | * Only apic need an MMIO device hook, so shortcut now.. | |
1067 | */ | |
1068 | static struct kvm_io_device *vcpu_find_pervcpu_dev(struct kvm_vcpu *vcpu, | |
1069 | gpa_t addr) | |
1070 | { | |
1071 | struct kvm_io_device *dev; | |
1072 | ||
1073 | if (vcpu->apic) { | |
1074 | dev = &vcpu->apic->dev; | |
1075 | if (dev->in_range(dev, addr)) | |
1076 | return dev; | |
1077 | } | |
1078 | return NULL; | |
1079 | } | |
1080 | ||
2eeb2e94 GH |
1081 | static struct kvm_io_device *vcpu_find_mmio_dev(struct kvm_vcpu *vcpu, |
1082 | gpa_t addr) | |
1083 | { | |
97222cc8 ED |
1084 | struct kvm_io_device *dev; |
1085 | ||
1086 | dev = vcpu_find_pervcpu_dev(vcpu, addr); | |
1087 | if (dev == NULL) | |
1088 | dev = kvm_io_bus_find_dev(&vcpu->kvm->mmio_bus, addr); | |
1089 | return dev; | |
2eeb2e94 GH |
1090 | } |
1091 | ||
74906345 ED |
1092 | static struct kvm_io_device *vcpu_find_pio_dev(struct kvm_vcpu *vcpu, |
1093 | gpa_t addr) | |
1094 | { | |
1095 | return kvm_io_bus_find_dev(&vcpu->kvm->pio_bus, addr); | |
1096 | } | |
1097 | ||
6aa8b732 | 1098 | static int emulator_read_emulated(unsigned long addr, |
4c690a1e | 1099 | void *val, |
6aa8b732 | 1100 | unsigned int bytes, |
cebff02b | 1101 | struct kvm_vcpu *vcpu) |
6aa8b732 | 1102 | { |
2eeb2e94 GH |
1103 | struct kvm_io_device *mmio_dev; |
1104 | gpa_t gpa; | |
6aa8b732 AK |
1105 | |
1106 | if (vcpu->mmio_read_completed) { | |
1107 | memcpy(val, vcpu->mmio_data, bytes); | |
1108 | vcpu->mmio_read_completed = 0; | |
1109 | return X86EMUL_CONTINUE; | |
cebff02b | 1110 | } else if (emulator_read_std(addr, val, bytes, vcpu) |
6aa8b732 AK |
1111 | == X86EMUL_CONTINUE) |
1112 | return X86EMUL_CONTINUE; | |
d27d4aca | 1113 | |
2eeb2e94 GH |
1114 | gpa = vcpu->mmu.gva_to_gpa(vcpu, addr); |
1115 | if (gpa == UNMAPPED_GVA) | |
1116 | return X86EMUL_PROPAGATE_FAULT; | |
6aa8b732 | 1117 | |
2eeb2e94 GH |
1118 | /* |
1119 | * Is this MMIO handled locally? | |
1120 | */ | |
1121 | mmio_dev = vcpu_find_mmio_dev(vcpu, gpa); | |
1122 | if (mmio_dev) { | |
1123 | kvm_iodevice_read(mmio_dev, gpa, bytes, val); | |
1124 | return X86EMUL_CONTINUE; | |
6aa8b732 | 1125 | } |
2eeb2e94 GH |
1126 | |
1127 | vcpu->mmio_needed = 1; | |
1128 | vcpu->mmio_phys_addr = gpa; | |
1129 | vcpu->mmio_size = bytes; | |
1130 | vcpu->mmio_is_write = 0; | |
1131 | ||
1132 | return X86EMUL_UNHANDLEABLE; | |
6aa8b732 AK |
1133 | } |
1134 | ||
da4a00f0 | 1135 | static int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa, |
4c690a1e | 1136 | const void *val, int bytes) |
da4a00f0 | 1137 | { |
da4a00f0 AK |
1138 | struct page *page; |
1139 | void *virt; | |
1140 | ||
1141 | if (((gpa + bytes - 1) >> PAGE_SHIFT) != (gpa >> PAGE_SHIFT)) | |
1142 | return 0; | |
954bbbc2 AK |
1143 | page = gfn_to_page(vcpu->kvm, gpa >> PAGE_SHIFT); |
1144 | if (!page) | |
da4a00f0 | 1145 | return 0; |
ab51a434 | 1146 | mark_page_dirty(vcpu->kvm, gpa >> PAGE_SHIFT); |
da4a00f0 | 1147 | virt = kmap_atomic(page, KM_USER0); |
fe551881 | 1148 | kvm_mmu_pte_write(vcpu, gpa, val, bytes); |
7cfa4b0a | 1149 | memcpy(virt + offset_in_page(gpa), val, bytes); |
da4a00f0 | 1150 | kunmap_atomic(virt, KM_USER0); |
da4a00f0 AK |
1151 | return 1; |
1152 | } | |
1153 | ||
b0fcd903 AK |
1154 | static int emulator_write_emulated_onepage(unsigned long addr, |
1155 | const void *val, | |
1156 | unsigned int bytes, | |
cebff02b | 1157 | struct kvm_vcpu *vcpu) |
6aa8b732 | 1158 | { |
2eeb2e94 GH |
1159 | struct kvm_io_device *mmio_dev; |
1160 | gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr); | |
6aa8b732 | 1161 | |
c9047f53 AK |
1162 | if (gpa == UNMAPPED_GVA) { |
1163 | kvm_arch_ops->inject_page_fault(vcpu, addr, 2); | |
6aa8b732 | 1164 | return X86EMUL_PROPAGATE_FAULT; |
c9047f53 | 1165 | } |
6aa8b732 | 1166 | |
da4a00f0 AK |
1167 | if (emulator_write_phys(vcpu, gpa, val, bytes)) |
1168 | return X86EMUL_CONTINUE; | |
1169 | ||
2eeb2e94 GH |
1170 | /* |
1171 | * Is this MMIO handled locally? | |
1172 | */ | |
1173 | mmio_dev = vcpu_find_mmio_dev(vcpu, gpa); | |
1174 | if (mmio_dev) { | |
1175 | kvm_iodevice_write(mmio_dev, gpa, bytes, val); | |
1176 | return X86EMUL_CONTINUE; | |
1177 | } | |
1178 | ||
6aa8b732 AK |
1179 | vcpu->mmio_needed = 1; |
1180 | vcpu->mmio_phys_addr = gpa; | |
1181 | vcpu->mmio_size = bytes; | |
1182 | vcpu->mmio_is_write = 1; | |
4c690a1e | 1183 | memcpy(vcpu->mmio_data, val, bytes); |
6aa8b732 AK |
1184 | |
1185 | return X86EMUL_CONTINUE; | |
1186 | } | |
1187 | ||
e7d5d76c | 1188 | int emulator_write_emulated(unsigned long addr, |
b0fcd903 AK |
1189 | const void *val, |
1190 | unsigned int bytes, | |
cebff02b | 1191 | struct kvm_vcpu *vcpu) |
b0fcd903 AK |
1192 | { |
1193 | /* Crossing a page boundary? */ | |
1194 | if (((addr + bytes - 1) ^ addr) & PAGE_MASK) { | |
1195 | int rc, now; | |
1196 | ||
1197 | now = -addr & ~PAGE_MASK; | |
cebff02b | 1198 | rc = emulator_write_emulated_onepage(addr, val, now, vcpu); |
b0fcd903 AK |
1199 | if (rc != X86EMUL_CONTINUE) |
1200 | return rc; | |
1201 | addr += now; | |
1202 | val += now; | |
1203 | bytes -= now; | |
1204 | } | |
cebff02b | 1205 | return emulator_write_emulated_onepage(addr, val, bytes, vcpu); |
b0fcd903 | 1206 | } |
e7d5d76c | 1207 | EXPORT_SYMBOL_GPL(emulator_write_emulated); |
b0fcd903 | 1208 | |
6aa8b732 | 1209 | static int emulator_cmpxchg_emulated(unsigned long addr, |
4c690a1e AK |
1210 | const void *old, |
1211 | const void *new, | |
6aa8b732 | 1212 | unsigned int bytes, |
cebff02b | 1213 | struct kvm_vcpu *vcpu) |
6aa8b732 AK |
1214 | { |
1215 | static int reported; | |
1216 | ||
1217 | if (!reported) { | |
1218 | reported = 1; | |
1219 | printk(KERN_WARNING "kvm: emulating exchange as write\n"); | |
1220 | } | |
cebff02b | 1221 | return emulator_write_emulated(addr, new, bytes, vcpu); |
6aa8b732 AK |
1222 | } |
1223 | ||
1224 | static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg) | |
1225 | { | |
1226 | return kvm_arch_ops->get_segment_base(vcpu, seg); | |
1227 | } | |
1228 | ||
1229 | int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address) | |
1230 | { | |
6aa8b732 AK |
1231 | return X86EMUL_CONTINUE; |
1232 | } | |
1233 | ||
1234 | int emulate_clts(struct kvm_vcpu *vcpu) | |
1235 | { | |
399badf3 | 1236 | unsigned long cr0; |
6aa8b732 | 1237 | |
707d92fa | 1238 | cr0 = vcpu->cr0 & ~X86_CR0_TS; |
6aa8b732 AK |
1239 | kvm_arch_ops->set_cr0(vcpu, cr0); |
1240 | return X86EMUL_CONTINUE; | |
1241 | } | |
1242 | ||
1243 | int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr, unsigned long *dest) | |
1244 | { | |
1245 | struct kvm_vcpu *vcpu = ctxt->vcpu; | |
1246 | ||
1247 | switch (dr) { | |
1248 | case 0 ... 3: | |
1249 | *dest = kvm_arch_ops->get_dr(vcpu, dr); | |
1250 | return X86EMUL_CONTINUE; | |
1251 | default: | |
f0242478 | 1252 | pr_unimpl(vcpu, "%s: unexpected dr %u\n", __FUNCTION__, dr); |
6aa8b732 AK |
1253 | return X86EMUL_UNHANDLEABLE; |
1254 | } | |
1255 | } | |
1256 | ||
1257 | int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr, unsigned long value) | |
1258 | { | |
1259 | unsigned long mask = (ctxt->mode == X86EMUL_MODE_PROT64) ? ~0ULL : ~0U; | |
1260 | int exception; | |
1261 | ||
1262 | kvm_arch_ops->set_dr(ctxt->vcpu, dr, value & mask, &exception); | |
1263 | if (exception) { | |
1264 | /* FIXME: better handling */ | |
1265 | return X86EMUL_UNHANDLEABLE; | |
1266 | } | |
1267 | return X86EMUL_CONTINUE; | |
1268 | } | |
1269 | ||
1270 | static void report_emulation_failure(struct x86_emulate_ctxt *ctxt) | |
1271 | { | |
1272 | static int reported; | |
1273 | u8 opcodes[4]; | |
1274 | unsigned long rip = ctxt->vcpu->rip; | |
1275 | unsigned long rip_linear; | |
1276 | ||
1277 | rip_linear = rip + get_segment_base(ctxt->vcpu, VCPU_SREG_CS); | |
1278 | ||
1279 | if (reported) | |
1280 | return; | |
1281 | ||
cebff02b | 1282 | emulator_read_std(rip_linear, (void *)opcodes, 4, ctxt->vcpu); |
6aa8b732 AK |
1283 | |
1284 | printk(KERN_ERR "emulation failed but !mmio_needed?" | |
1285 | " rip %lx %02x %02x %02x %02x\n", | |
1286 | rip, opcodes[0], opcodes[1], opcodes[2], opcodes[3]); | |
1287 | reported = 1; | |
1288 | } | |
1289 | ||
1290 | struct x86_emulate_ops emulate_ops = { | |
1291 | .read_std = emulator_read_std, | |
1292 | .write_std = emulator_write_std, | |
1293 | .read_emulated = emulator_read_emulated, | |
1294 | .write_emulated = emulator_write_emulated, | |
1295 | .cmpxchg_emulated = emulator_cmpxchg_emulated, | |
1296 | }; | |
1297 | ||
1298 | int emulate_instruction(struct kvm_vcpu *vcpu, | |
1299 | struct kvm_run *run, | |
1300 | unsigned long cr2, | |
1301 | u16 error_code) | |
1302 | { | |
1303 | struct x86_emulate_ctxt emulate_ctxt; | |
1304 | int r; | |
1305 | int cs_db, cs_l; | |
1306 | ||
e7df56e4 | 1307 | vcpu->mmio_fault_cr2 = cr2; |
6aa8b732 AK |
1308 | kvm_arch_ops->cache_regs(vcpu); |
1309 | ||
1310 | kvm_arch_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l); | |
1311 | ||
1312 | emulate_ctxt.vcpu = vcpu; | |
1313 | emulate_ctxt.eflags = kvm_arch_ops->get_rflags(vcpu); | |
1314 | emulate_ctxt.cr2 = cr2; | |
1315 | emulate_ctxt.mode = (emulate_ctxt.eflags & X86_EFLAGS_VM) | |
1316 | ? X86EMUL_MODE_REAL : cs_l | |
1317 | ? X86EMUL_MODE_PROT64 : cs_db | |
1318 | ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16; | |
1319 | ||
1320 | if (emulate_ctxt.mode == X86EMUL_MODE_PROT64) { | |
1321 | emulate_ctxt.cs_base = 0; | |
1322 | emulate_ctxt.ds_base = 0; | |
1323 | emulate_ctxt.es_base = 0; | |
1324 | emulate_ctxt.ss_base = 0; | |
1325 | } else { | |
1326 | emulate_ctxt.cs_base = get_segment_base(vcpu, VCPU_SREG_CS); | |
1327 | emulate_ctxt.ds_base = get_segment_base(vcpu, VCPU_SREG_DS); | |
1328 | emulate_ctxt.es_base = get_segment_base(vcpu, VCPU_SREG_ES); | |
1329 | emulate_ctxt.ss_base = get_segment_base(vcpu, VCPU_SREG_SS); | |
1330 | } | |
1331 | ||
1332 | emulate_ctxt.gs_base = get_segment_base(vcpu, VCPU_SREG_GS); | |
1333 | emulate_ctxt.fs_base = get_segment_base(vcpu, VCPU_SREG_FS); | |
1334 | ||
1335 | vcpu->mmio_is_write = 0; | |
e70669ab | 1336 | vcpu->pio.string = 0; |
6aa8b732 | 1337 | r = x86_emulate_memop(&emulate_ctxt, &emulate_ops); |
e70669ab LV |
1338 | if (vcpu->pio.string) |
1339 | return EMULATE_DO_MMIO; | |
6aa8b732 AK |
1340 | |
1341 | if ((r || vcpu->mmio_is_write) && run) { | |
8fc0d085 | 1342 | run->exit_reason = KVM_EXIT_MMIO; |
6aa8b732 AK |
1343 | run->mmio.phys_addr = vcpu->mmio_phys_addr; |
1344 | memcpy(run->mmio.data, vcpu->mmio_data, 8); | |
1345 | run->mmio.len = vcpu->mmio_size; | |
1346 | run->mmio.is_write = vcpu->mmio_is_write; | |
1347 | } | |
1348 | ||
1349 | if (r) { | |
a436036b AK |
1350 | if (kvm_mmu_unprotect_page_virt(vcpu, cr2)) |
1351 | return EMULATE_DONE; | |
6aa8b732 AK |
1352 | if (!vcpu->mmio_needed) { |
1353 | report_emulation_failure(&emulate_ctxt); | |
1354 | return EMULATE_FAIL; | |
1355 | } | |
1356 | return EMULATE_DO_MMIO; | |
1357 | } | |
1358 | ||
1359 | kvm_arch_ops->decache_regs(vcpu); | |
1360 | kvm_arch_ops->set_rflags(vcpu, emulate_ctxt.eflags); | |
1361 | ||
02c83209 AK |
1362 | if (vcpu->mmio_is_write) { |
1363 | vcpu->mmio_needed = 0; | |
6aa8b732 | 1364 | return EMULATE_DO_MMIO; |
02c83209 | 1365 | } |
6aa8b732 AK |
1366 | |
1367 | return EMULATE_DONE; | |
1368 | } | |
1369 | EXPORT_SYMBOL_GPL(emulate_instruction); | |
1370 | ||
b6958ce4 ED |
1371 | /* |
1372 | * The vCPU has executed a HLT instruction with in-kernel mode enabled. | |
1373 | */ | |
1374 | static void kvm_vcpu_kernel_halt(struct kvm_vcpu *vcpu) | |
d3bef15f | 1375 | { |
b6958ce4 ED |
1376 | DECLARE_WAITQUEUE(wait, current); |
1377 | ||
1378 | add_wait_queue(&vcpu->wq, &wait); | |
1379 | ||
1380 | /* | |
1381 | * We will block until either an interrupt or a signal wakes us up | |
1382 | */ | |
1383 | while(!(irqchip_in_kernel(vcpu->kvm) && kvm_cpu_has_interrupt(vcpu)) | |
1384 | && !vcpu->irq_summary | |
1385 | && !signal_pending(current)) { | |
1386 | set_current_state(TASK_INTERRUPTIBLE); | |
1387 | vcpu_put(vcpu); | |
1388 | schedule(); | |
1389 | vcpu_load(vcpu); | |
1390 | } | |
d3bef15f | 1391 | |
b6958ce4 ED |
1392 | remove_wait_queue(&vcpu->wq, &wait); |
1393 | set_current_state(TASK_RUNNING); | |
1394 | } | |
1395 | ||
1396 | int kvm_emulate_halt(struct kvm_vcpu *vcpu) | |
1397 | { | |
d3bef15f | 1398 | ++vcpu->stat.halt_exits; |
b6958ce4 ED |
1399 | if (irqchip_in_kernel(vcpu->kvm)) { |
1400 | kvm_vcpu_kernel_halt(vcpu); | |
1401 | return 1; | |
1402 | } else { | |
1403 | vcpu->run->exit_reason = KVM_EXIT_HLT; | |
1404 | return 0; | |
1405 | } | |
d3bef15f AK |
1406 | } |
1407 | EXPORT_SYMBOL_GPL(kvm_emulate_halt); | |
1408 | ||
270fd9b9 AK |
1409 | int kvm_hypercall(struct kvm_vcpu *vcpu, struct kvm_run *run) |
1410 | { | |
1411 | unsigned long nr, a0, a1, a2, a3, a4, a5, ret; | |
1412 | ||
9b22bf57 | 1413 | kvm_arch_ops->cache_regs(vcpu); |
270fd9b9 AK |
1414 | ret = -KVM_EINVAL; |
1415 | #ifdef CONFIG_X86_64 | |
1416 | if (is_long_mode(vcpu)) { | |
1417 | nr = vcpu->regs[VCPU_REGS_RAX]; | |
1418 | a0 = vcpu->regs[VCPU_REGS_RDI]; | |
1419 | a1 = vcpu->regs[VCPU_REGS_RSI]; | |
1420 | a2 = vcpu->regs[VCPU_REGS_RDX]; | |
1421 | a3 = vcpu->regs[VCPU_REGS_RCX]; | |
1422 | a4 = vcpu->regs[VCPU_REGS_R8]; | |
1423 | a5 = vcpu->regs[VCPU_REGS_R9]; | |
1424 | } else | |
1425 | #endif | |
1426 | { | |
1427 | nr = vcpu->regs[VCPU_REGS_RBX] & -1u; | |
1428 | a0 = vcpu->regs[VCPU_REGS_RAX] & -1u; | |
1429 | a1 = vcpu->regs[VCPU_REGS_RCX] & -1u; | |
1430 | a2 = vcpu->regs[VCPU_REGS_RDX] & -1u; | |
1431 | a3 = vcpu->regs[VCPU_REGS_RSI] & -1u; | |
1432 | a4 = vcpu->regs[VCPU_REGS_RDI] & -1u; | |
1433 | a5 = vcpu->regs[VCPU_REGS_RBP] & -1u; | |
1434 | } | |
1435 | switch (nr) { | |
1436 | default: | |
519ef353 | 1437 | run->hypercall.nr = nr; |
b4e63f56 AK |
1438 | run->hypercall.args[0] = a0; |
1439 | run->hypercall.args[1] = a1; | |
1440 | run->hypercall.args[2] = a2; | |
1441 | run->hypercall.args[3] = a3; | |
1442 | run->hypercall.args[4] = a4; | |
1443 | run->hypercall.args[5] = a5; | |
1444 | run->hypercall.ret = ret; | |
1445 | run->hypercall.longmode = is_long_mode(vcpu); | |
1446 | kvm_arch_ops->decache_regs(vcpu); | |
1447 | return 0; | |
270fd9b9 AK |
1448 | } |
1449 | vcpu->regs[VCPU_REGS_RAX] = ret; | |
9b22bf57 | 1450 | kvm_arch_ops->decache_regs(vcpu); |
270fd9b9 AK |
1451 | return 1; |
1452 | } | |
1453 | EXPORT_SYMBOL_GPL(kvm_hypercall); | |
1454 | ||
6aa8b732 AK |
1455 | static u64 mk_cr_64(u64 curr_cr, u32 new_val) |
1456 | { | |
1457 | return (curr_cr & ~((1ULL << 32) - 1)) | new_val; | |
1458 | } | |
1459 | ||
1460 | void realmode_lgdt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base) | |
1461 | { | |
1462 | struct descriptor_table dt = { limit, base }; | |
1463 | ||
1464 | kvm_arch_ops->set_gdt(vcpu, &dt); | |
1465 | } | |
1466 | ||
1467 | void realmode_lidt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base) | |
1468 | { | |
1469 | struct descriptor_table dt = { limit, base }; | |
1470 | ||
1471 | kvm_arch_ops->set_idt(vcpu, &dt); | |
1472 | } | |
1473 | ||
1474 | void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw, | |
1475 | unsigned long *rflags) | |
1476 | { | |
1477 | lmsw(vcpu, msw); | |
1478 | *rflags = kvm_arch_ops->get_rflags(vcpu); | |
1479 | } | |
1480 | ||
1481 | unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr) | |
1482 | { | |
25c4c276 | 1483 | kvm_arch_ops->decache_cr4_guest_bits(vcpu); |
6aa8b732 AK |
1484 | switch (cr) { |
1485 | case 0: | |
1486 | return vcpu->cr0; | |
1487 | case 2: | |
1488 | return vcpu->cr2; | |
1489 | case 3: | |
1490 | return vcpu->cr3; | |
1491 | case 4: | |
1492 | return vcpu->cr4; | |
1493 | default: | |
1494 | vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr); | |
1495 | return 0; | |
1496 | } | |
1497 | } | |
1498 | ||
1499 | void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long val, | |
1500 | unsigned long *rflags) | |
1501 | { | |
1502 | switch (cr) { | |
1503 | case 0: | |
1504 | set_cr0(vcpu, mk_cr_64(vcpu->cr0, val)); | |
1505 | *rflags = kvm_arch_ops->get_rflags(vcpu); | |
1506 | break; | |
1507 | case 2: | |
1508 | vcpu->cr2 = val; | |
1509 | break; | |
1510 | case 3: | |
1511 | set_cr3(vcpu, val); | |
1512 | break; | |
1513 | case 4: | |
1514 | set_cr4(vcpu, mk_cr_64(vcpu->cr4, val)); | |
1515 | break; | |
1516 | default: | |
1517 | vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr); | |
1518 | } | |
1519 | } | |
1520 | ||
102d8325 IM |
1521 | /* |
1522 | * Register the para guest with the host: | |
1523 | */ | |
1524 | static int vcpu_register_para(struct kvm_vcpu *vcpu, gpa_t para_state_gpa) | |
1525 | { | |
1526 | struct kvm_vcpu_para_state *para_state; | |
1527 | hpa_t para_state_hpa, hypercall_hpa; | |
1528 | struct page *para_state_page; | |
1529 | unsigned char *hypercall; | |
1530 | gpa_t hypercall_gpa; | |
1531 | ||
1532 | printk(KERN_DEBUG "kvm: guest trying to enter paravirtual mode\n"); | |
1533 | printk(KERN_DEBUG ".... para_state_gpa: %08Lx\n", para_state_gpa); | |
1534 | ||
1535 | /* | |
1536 | * Needs to be page aligned: | |
1537 | */ | |
1538 | if (para_state_gpa != PAGE_ALIGN(para_state_gpa)) | |
1539 | goto err_gp; | |
1540 | ||
1541 | para_state_hpa = gpa_to_hpa(vcpu, para_state_gpa); | |
1542 | printk(KERN_DEBUG ".... para_state_hpa: %08Lx\n", para_state_hpa); | |
1543 | if (is_error_hpa(para_state_hpa)) | |
1544 | goto err_gp; | |
1545 | ||
ab51a434 | 1546 | mark_page_dirty(vcpu->kvm, para_state_gpa >> PAGE_SHIFT); |
102d8325 | 1547 | para_state_page = pfn_to_page(para_state_hpa >> PAGE_SHIFT); |
fe551881 | 1548 | para_state = kmap(para_state_page); |
102d8325 IM |
1549 | |
1550 | printk(KERN_DEBUG ".... guest version: %d\n", para_state->guest_version); | |
1551 | printk(KERN_DEBUG ".... size: %d\n", para_state->size); | |
1552 | ||
1553 | para_state->host_version = KVM_PARA_API_VERSION; | |
1554 | /* | |
1555 | * We cannot support guests that try to register themselves | |
1556 | * with a newer API version than the host supports: | |
1557 | */ | |
1558 | if (para_state->guest_version > KVM_PARA_API_VERSION) { | |
1559 | para_state->ret = -KVM_EINVAL; | |
1560 | goto err_kunmap_skip; | |
1561 | } | |
1562 | ||
1563 | hypercall_gpa = para_state->hypercall_gpa; | |
1564 | hypercall_hpa = gpa_to_hpa(vcpu, hypercall_gpa); | |
1565 | printk(KERN_DEBUG ".... hypercall_hpa: %08Lx\n", hypercall_hpa); | |
1566 | if (is_error_hpa(hypercall_hpa)) { | |
1567 | para_state->ret = -KVM_EINVAL; | |
1568 | goto err_kunmap_skip; | |
1569 | } | |
1570 | ||
1571 | printk(KERN_DEBUG "kvm: para guest successfully registered.\n"); | |
1572 | vcpu->para_state_page = para_state_page; | |
1573 | vcpu->para_state_gpa = para_state_gpa; | |
1574 | vcpu->hypercall_gpa = hypercall_gpa; | |
1575 | ||
ab51a434 | 1576 | mark_page_dirty(vcpu->kvm, hypercall_gpa >> PAGE_SHIFT); |
102d8325 IM |
1577 | hypercall = kmap_atomic(pfn_to_page(hypercall_hpa >> PAGE_SHIFT), |
1578 | KM_USER1) + (hypercall_hpa & ~PAGE_MASK); | |
1579 | kvm_arch_ops->patch_hypercall(vcpu, hypercall); | |
1580 | kunmap_atomic(hypercall, KM_USER1); | |
1581 | ||
1582 | para_state->ret = 0; | |
1583 | err_kunmap_skip: | |
fe551881 | 1584 | kunmap(para_state_page); |
102d8325 IM |
1585 | return 0; |
1586 | err_gp: | |
1587 | return 1; | |
1588 | } | |
1589 | ||
3bab1f5d AK |
1590 | int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata) |
1591 | { | |
1592 | u64 data; | |
1593 | ||
1594 | switch (msr) { | |
1595 | case 0xc0010010: /* SYSCFG */ | |
1596 | case 0xc0010015: /* HWCR */ | |
1597 | case MSR_IA32_PLATFORM_ID: | |
1598 | case MSR_IA32_P5_MC_ADDR: | |
1599 | case MSR_IA32_P5_MC_TYPE: | |
1600 | case MSR_IA32_MC0_CTL: | |
1601 | case MSR_IA32_MCG_STATUS: | |
1602 | case MSR_IA32_MCG_CAP: | |
1603 | case MSR_IA32_MC0_MISC: | |
1604 | case MSR_IA32_MC0_MISC+4: | |
1605 | case MSR_IA32_MC0_MISC+8: | |
1606 | case MSR_IA32_MC0_MISC+12: | |
1607 | case MSR_IA32_MC0_MISC+16: | |
1608 | case MSR_IA32_UCODE_REV: | |
a8d13ea2 | 1609 | case MSR_IA32_PERF_STATUS: |
2dc7094b | 1610 | case MSR_IA32_EBL_CR_POWERON: |
3bab1f5d AK |
1611 | /* MTRR registers */ |
1612 | case 0xfe: | |
1613 | case 0x200 ... 0x2ff: | |
1614 | data = 0; | |
1615 | break; | |
a8d13ea2 AK |
1616 | case 0xcd: /* fsb frequency */ |
1617 | data = 3; | |
1618 | break; | |
3bab1f5d | 1619 | case MSR_IA32_APICBASE: |
7017fc3d | 1620 | data = kvm_get_apic_base(vcpu); |
3bab1f5d | 1621 | break; |
6f00e68f AK |
1622 | case MSR_IA32_MISC_ENABLE: |
1623 | data = vcpu->ia32_misc_enable_msr; | |
1624 | break; | |
3bab1f5d AK |
1625 | #ifdef CONFIG_X86_64 |
1626 | case MSR_EFER: | |
1627 | data = vcpu->shadow_efer; | |
1628 | break; | |
1629 | #endif | |
1630 | default: | |
f0242478 | 1631 | pr_unimpl(vcpu, "unhandled rdmsr: 0x%x\n", msr); |
3bab1f5d AK |
1632 | return 1; |
1633 | } | |
1634 | *pdata = data; | |
1635 | return 0; | |
1636 | } | |
1637 | EXPORT_SYMBOL_GPL(kvm_get_msr_common); | |
1638 | ||
6aa8b732 AK |
1639 | /* |
1640 | * Reads an msr value (of 'msr_index') into 'pdata'. | |
1641 | * Returns 0 on success, non-0 otherwise. | |
1642 | * Assumes vcpu_load() was already called. | |
1643 | */ | |
35f3f286 | 1644 | int kvm_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata) |
6aa8b732 AK |
1645 | { |
1646 | return kvm_arch_ops->get_msr(vcpu, msr_index, pdata); | |
1647 | } | |
1648 | ||
05b3e0c2 | 1649 | #ifdef CONFIG_X86_64 |
6aa8b732 | 1650 | |
3bab1f5d | 1651 | static void set_efer(struct kvm_vcpu *vcpu, u64 efer) |
6aa8b732 | 1652 | { |
6aa8b732 AK |
1653 | if (efer & EFER_RESERVED_BITS) { |
1654 | printk(KERN_DEBUG "set_efer: 0x%llx #GP, reserved bits\n", | |
1655 | efer); | |
1656 | inject_gp(vcpu); | |
1657 | return; | |
1658 | } | |
1659 | ||
1660 | if (is_paging(vcpu) | |
1661 | && (vcpu->shadow_efer & EFER_LME) != (efer & EFER_LME)) { | |
1662 | printk(KERN_DEBUG "set_efer: #GP, change LME while paging\n"); | |
1663 | inject_gp(vcpu); | |
1664 | return; | |
1665 | } | |
1666 | ||
7725f0ba AK |
1667 | kvm_arch_ops->set_efer(vcpu, efer); |
1668 | ||
6aa8b732 AK |
1669 | efer &= ~EFER_LMA; |
1670 | efer |= vcpu->shadow_efer & EFER_LMA; | |
1671 | ||
1672 | vcpu->shadow_efer = efer; | |
6aa8b732 | 1673 | } |
6aa8b732 AK |
1674 | |
1675 | #endif | |
1676 | ||
3bab1f5d AK |
1677 | int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data) |
1678 | { | |
1679 | switch (msr) { | |
1680 | #ifdef CONFIG_X86_64 | |
1681 | case MSR_EFER: | |
1682 | set_efer(vcpu, data); | |
1683 | break; | |
1684 | #endif | |
1685 | case MSR_IA32_MC0_STATUS: | |
f0242478 | 1686 | pr_unimpl(vcpu, "%s: MSR_IA32_MC0_STATUS 0x%llx, nop\n", |
3bab1f5d AK |
1687 | __FUNCTION__, data); |
1688 | break; | |
0e5bf0d0 | 1689 | case MSR_IA32_MCG_STATUS: |
f0242478 | 1690 | pr_unimpl(vcpu, "%s: MSR_IA32_MCG_STATUS 0x%llx, nop\n", |
0e5bf0d0 SK |
1691 | __FUNCTION__, data); |
1692 | break; | |
3bab1f5d AK |
1693 | case MSR_IA32_UCODE_REV: |
1694 | case MSR_IA32_UCODE_WRITE: | |
1695 | case 0x200 ... 0x2ff: /* MTRRs */ | |
1696 | break; | |
1697 | case MSR_IA32_APICBASE: | |
7017fc3d | 1698 | kvm_set_apic_base(vcpu, data); |
3bab1f5d | 1699 | break; |
6f00e68f AK |
1700 | case MSR_IA32_MISC_ENABLE: |
1701 | vcpu->ia32_misc_enable_msr = data; | |
1702 | break; | |
102d8325 IM |
1703 | /* |
1704 | * This is the 'probe whether the host is KVM' logic: | |
1705 | */ | |
1706 | case MSR_KVM_API_MAGIC: | |
1707 | return vcpu_register_para(vcpu, data); | |
1708 | ||
3bab1f5d | 1709 | default: |
f0242478 | 1710 | pr_unimpl(vcpu, "unhandled wrmsr: 0x%x\n", msr); |
3bab1f5d AK |
1711 | return 1; |
1712 | } | |
1713 | return 0; | |
1714 | } | |
1715 | EXPORT_SYMBOL_GPL(kvm_set_msr_common); | |
1716 | ||
6aa8b732 AK |
1717 | /* |
1718 | * Writes msr value into into the appropriate "register". | |
1719 | * Returns 0 on success, non-0 otherwise. | |
1720 | * Assumes vcpu_load() was already called. | |
1721 | */ | |
35f3f286 | 1722 | int kvm_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data) |
6aa8b732 AK |
1723 | { |
1724 | return kvm_arch_ops->set_msr(vcpu, msr_index, data); | |
1725 | } | |
1726 | ||
1727 | void kvm_resched(struct kvm_vcpu *vcpu) | |
1728 | { | |
3fca0365 YD |
1729 | if (!need_resched()) |
1730 | return; | |
6aa8b732 | 1731 | cond_resched(); |
6aa8b732 AK |
1732 | } |
1733 | EXPORT_SYMBOL_GPL(kvm_resched); | |
1734 | ||
06465c5a AK |
1735 | void kvm_emulate_cpuid(struct kvm_vcpu *vcpu) |
1736 | { | |
1737 | int i; | |
1738 | u32 function; | |
1739 | struct kvm_cpuid_entry *e, *best; | |
1740 | ||
1741 | kvm_arch_ops->cache_regs(vcpu); | |
1742 | function = vcpu->regs[VCPU_REGS_RAX]; | |
1743 | vcpu->regs[VCPU_REGS_RAX] = 0; | |
1744 | vcpu->regs[VCPU_REGS_RBX] = 0; | |
1745 | vcpu->regs[VCPU_REGS_RCX] = 0; | |
1746 | vcpu->regs[VCPU_REGS_RDX] = 0; | |
1747 | best = NULL; | |
1748 | for (i = 0; i < vcpu->cpuid_nent; ++i) { | |
1749 | e = &vcpu->cpuid_entries[i]; | |
1750 | if (e->function == function) { | |
1751 | best = e; | |
1752 | break; | |
1753 | } | |
1754 | /* | |
1755 | * Both basic or both extended? | |
1756 | */ | |
1757 | if (((e->function ^ function) & 0x80000000) == 0) | |
1758 | if (!best || e->function > best->function) | |
1759 | best = e; | |
1760 | } | |
1761 | if (best) { | |
1762 | vcpu->regs[VCPU_REGS_RAX] = best->eax; | |
1763 | vcpu->regs[VCPU_REGS_RBX] = best->ebx; | |
1764 | vcpu->regs[VCPU_REGS_RCX] = best->ecx; | |
1765 | vcpu->regs[VCPU_REGS_RDX] = best->edx; | |
1766 | } | |
1767 | kvm_arch_ops->decache_regs(vcpu); | |
1768 | kvm_arch_ops->skip_emulated_instruction(vcpu); | |
1769 | } | |
1770 | EXPORT_SYMBOL_GPL(kvm_emulate_cpuid); | |
1771 | ||
039576c0 | 1772 | static int pio_copy_data(struct kvm_vcpu *vcpu) |
46fc1477 | 1773 | { |
039576c0 AK |
1774 | void *p = vcpu->pio_data; |
1775 | void *q; | |
1776 | unsigned bytes; | |
1777 | int nr_pages = vcpu->pio.guest_pages[1] ? 2 : 1; | |
1778 | ||
039576c0 AK |
1779 | q = vmap(vcpu->pio.guest_pages, nr_pages, VM_READ|VM_WRITE, |
1780 | PAGE_KERNEL); | |
1781 | if (!q) { | |
039576c0 AK |
1782 | free_pio_guest_pages(vcpu); |
1783 | return -ENOMEM; | |
1784 | } | |
1785 | q += vcpu->pio.guest_page_offset; | |
1786 | bytes = vcpu->pio.size * vcpu->pio.cur_count; | |
1787 | if (vcpu->pio.in) | |
1788 | memcpy(q, p, bytes); | |
1789 | else | |
1790 | memcpy(p, q, bytes); | |
1791 | q -= vcpu->pio.guest_page_offset; | |
1792 | vunmap(q); | |
039576c0 AK |
1793 | free_pio_guest_pages(vcpu); |
1794 | return 0; | |
1795 | } | |
1796 | ||
1797 | static int complete_pio(struct kvm_vcpu *vcpu) | |
1798 | { | |
1799 | struct kvm_pio_request *io = &vcpu->pio; | |
46fc1477 | 1800 | long delta; |
039576c0 | 1801 | int r; |
46fc1477 AK |
1802 | |
1803 | kvm_arch_ops->cache_regs(vcpu); | |
1804 | ||
1805 | if (!io->string) { | |
039576c0 AK |
1806 | if (io->in) |
1807 | memcpy(&vcpu->regs[VCPU_REGS_RAX], vcpu->pio_data, | |
46fc1477 AK |
1808 | io->size); |
1809 | } else { | |
039576c0 AK |
1810 | if (io->in) { |
1811 | r = pio_copy_data(vcpu); | |
1812 | if (r) { | |
1813 | kvm_arch_ops->cache_regs(vcpu); | |
1814 | return r; | |
1815 | } | |
1816 | } | |
1817 | ||
46fc1477 AK |
1818 | delta = 1; |
1819 | if (io->rep) { | |
039576c0 | 1820 | delta *= io->cur_count; |
46fc1477 AK |
1821 | /* |
1822 | * The size of the register should really depend on | |
1823 | * current address size. | |
1824 | */ | |
1825 | vcpu->regs[VCPU_REGS_RCX] -= delta; | |
1826 | } | |
039576c0 | 1827 | if (io->down) |
46fc1477 AK |
1828 | delta = -delta; |
1829 | delta *= io->size; | |
039576c0 | 1830 | if (io->in) |
46fc1477 AK |
1831 | vcpu->regs[VCPU_REGS_RDI] += delta; |
1832 | else | |
1833 | vcpu->regs[VCPU_REGS_RSI] += delta; | |
1834 | } | |
1835 | ||
46fc1477 AK |
1836 | kvm_arch_ops->decache_regs(vcpu); |
1837 | ||
039576c0 AK |
1838 | io->count -= io->cur_count; |
1839 | io->cur_count = 0; | |
1840 | ||
1841 | if (!io->count) | |
1842 | kvm_arch_ops->skip_emulated_instruction(vcpu); | |
1843 | return 0; | |
46fc1477 AK |
1844 | } |
1845 | ||
65619eb5 ED |
1846 | static void kernel_pio(struct kvm_io_device *pio_dev, |
1847 | struct kvm_vcpu *vcpu, | |
1848 | void *pd) | |
74906345 ED |
1849 | { |
1850 | /* TODO: String I/O for in kernel device */ | |
1851 | ||
9cf98828 | 1852 | mutex_lock(&vcpu->kvm->lock); |
74906345 ED |
1853 | if (vcpu->pio.in) |
1854 | kvm_iodevice_read(pio_dev, vcpu->pio.port, | |
1855 | vcpu->pio.size, | |
65619eb5 | 1856 | pd); |
74906345 ED |
1857 | else |
1858 | kvm_iodevice_write(pio_dev, vcpu->pio.port, | |
1859 | vcpu->pio.size, | |
65619eb5 | 1860 | pd); |
9cf98828 | 1861 | mutex_unlock(&vcpu->kvm->lock); |
65619eb5 ED |
1862 | } |
1863 | ||
1864 | static void pio_string_write(struct kvm_io_device *pio_dev, | |
1865 | struct kvm_vcpu *vcpu) | |
1866 | { | |
1867 | struct kvm_pio_request *io = &vcpu->pio; | |
1868 | void *pd = vcpu->pio_data; | |
1869 | int i; | |
1870 | ||
9cf98828 | 1871 | mutex_lock(&vcpu->kvm->lock); |
65619eb5 ED |
1872 | for (i = 0; i < io->cur_count; i++) { |
1873 | kvm_iodevice_write(pio_dev, io->port, | |
1874 | io->size, | |
1875 | pd); | |
1876 | pd += io->size; | |
1877 | } | |
9cf98828 | 1878 | mutex_unlock(&vcpu->kvm->lock); |
74906345 ED |
1879 | } |
1880 | ||
3090dd73 LV |
1881 | int kvm_emulate_pio (struct kvm_vcpu *vcpu, struct kvm_run *run, int in, |
1882 | int size, unsigned port) | |
1883 | { | |
1884 | struct kvm_io_device *pio_dev; | |
1885 | ||
1886 | vcpu->run->exit_reason = KVM_EXIT_IO; | |
1887 | vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; | |
1888 | vcpu->run->io.size = vcpu->pio.size = size; | |
1889 | vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; | |
1890 | vcpu->run->io.count = vcpu->pio.count = vcpu->pio.cur_count = 1; | |
1891 | vcpu->run->io.port = vcpu->pio.port = port; | |
1892 | vcpu->pio.in = in; | |
1893 | vcpu->pio.string = 0; | |
1894 | vcpu->pio.down = 0; | |
1895 | vcpu->pio.guest_page_offset = 0; | |
1896 | vcpu->pio.rep = 0; | |
1897 | ||
1898 | kvm_arch_ops->cache_regs(vcpu); | |
1899 | memcpy(vcpu->pio_data, &vcpu->regs[VCPU_REGS_RAX], 4); | |
1900 | kvm_arch_ops->decache_regs(vcpu); | |
1901 | ||
1902 | pio_dev = vcpu_find_pio_dev(vcpu, port); | |
1903 | if (pio_dev) { | |
1904 | kernel_pio(pio_dev, vcpu, vcpu->pio_data); | |
1905 | complete_pio(vcpu); | |
1906 | return 1; | |
1907 | } | |
1908 | return 0; | |
1909 | } | |
1910 | EXPORT_SYMBOL_GPL(kvm_emulate_pio); | |
1911 | ||
1912 | int kvm_emulate_pio_string(struct kvm_vcpu *vcpu, struct kvm_run *run, int in, | |
1913 | int size, unsigned long count, int down, | |
039576c0 AK |
1914 | gva_t address, int rep, unsigned port) |
1915 | { | |
1916 | unsigned now, in_page; | |
65619eb5 | 1917 | int i, ret = 0; |
039576c0 AK |
1918 | int nr_pages = 1; |
1919 | struct page *page; | |
74906345 | 1920 | struct kvm_io_device *pio_dev; |
039576c0 AK |
1921 | |
1922 | vcpu->run->exit_reason = KVM_EXIT_IO; | |
1923 | vcpu->run->io.direction = in ? KVM_EXIT_IO_IN : KVM_EXIT_IO_OUT; | |
3090dd73 | 1924 | vcpu->run->io.size = vcpu->pio.size = size; |
039576c0 | 1925 | vcpu->run->io.data_offset = KVM_PIO_PAGE_OFFSET * PAGE_SIZE; |
3090dd73 LV |
1926 | vcpu->run->io.count = vcpu->pio.count = vcpu->pio.cur_count = count; |
1927 | vcpu->run->io.port = vcpu->pio.port = port; | |
039576c0 | 1928 | vcpu->pio.in = in; |
3090dd73 | 1929 | vcpu->pio.string = 1; |
039576c0 AK |
1930 | vcpu->pio.down = down; |
1931 | vcpu->pio.guest_page_offset = offset_in_page(address); | |
1932 | vcpu->pio.rep = rep; | |
1933 | ||
039576c0 AK |
1934 | if (!count) { |
1935 | kvm_arch_ops->skip_emulated_instruction(vcpu); | |
1936 | return 1; | |
1937 | } | |
1938 | ||
039576c0 AK |
1939 | if (!down) |
1940 | in_page = PAGE_SIZE - offset_in_page(address); | |
1941 | else | |
1942 | in_page = offset_in_page(address) + size; | |
1943 | now = min(count, (unsigned long)in_page / size); | |
1944 | if (!now) { | |
1945 | /* | |
1946 | * String I/O straddles page boundary. Pin two guest pages | |
1947 | * so that we satisfy atomicity constraints. Do just one | |
1948 | * transaction to avoid complexity. | |
1949 | */ | |
1950 | nr_pages = 2; | |
1951 | now = 1; | |
1952 | } | |
1953 | if (down) { | |
1954 | /* | |
1955 | * String I/O in reverse. Yuck. Kill the guest, fix later. | |
1956 | */ | |
f0242478 | 1957 | pr_unimpl(vcpu, "guest string pio down\n"); |
039576c0 AK |
1958 | inject_gp(vcpu); |
1959 | return 1; | |
1960 | } | |
1961 | vcpu->run->io.count = now; | |
1962 | vcpu->pio.cur_count = now; | |
1963 | ||
1964 | for (i = 0; i < nr_pages; ++i) { | |
11ec2804 | 1965 | mutex_lock(&vcpu->kvm->lock); |
039576c0 AK |
1966 | page = gva_to_page(vcpu, address + i * PAGE_SIZE); |
1967 | if (page) | |
1968 | get_page(page); | |
1969 | vcpu->pio.guest_pages[i] = page; | |
11ec2804 | 1970 | mutex_unlock(&vcpu->kvm->lock); |
039576c0 AK |
1971 | if (!page) { |
1972 | inject_gp(vcpu); | |
1973 | free_pio_guest_pages(vcpu); | |
1974 | return 1; | |
1975 | } | |
1976 | } | |
1977 | ||
3090dd73 | 1978 | pio_dev = vcpu_find_pio_dev(vcpu, port); |
65619eb5 ED |
1979 | if (!vcpu->pio.in) { |
1980 | /* string PIO write */ | |
1981 | ret = pio_copy_data(vcpu); | |
1982 | if (ret >= 0 && pio_dev) { | |
1983 | pio_string_write(pio_dev, vcpu); | |
1984 | complete_pio(vcpu); | |
1985 | if (vcpu->pio.count == 0) | |
1986 | ret = 1; | |
1987 | } | |
1988 | } else if (pio_dev) | |
f0242478 | 1989 | pr_unimpl(vcpu, "no string pio read support yet, " |
65619eb5 ED |
1990 | "port %x size %d count %ld\n", |
1991 | port, size, count); | |
1992 | ||
1993 | return ret; | |
039576c0 | 1994 | } |
3090dd73 | 1995 | EXPORT_SYMBOL_GPL(kvm_emulate_pio_string); |
039576c0 | 1996 | |
bccf2150 | 1997 | static int kvm_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run) |
6aa8b732 | 1998 | { |
6aa8b732 | 1999 | int r; |
1961d276 | 2000 | sigset_t sigsaved; |
6aa8b732 | 2001 | |
bccf2150 | 2002 | vcpu_load(vcpu); |
6aa8b732 | 2003 | |
1961d276 AK |
2004 | if (vcpu->sigset_active) |
2005 | sigprocmask(SIG_SETMASK, &vcpu->sigset, &sigsaved); | |
2006 | ||
54810342 | 2007 | /* re-sync apic's tpr */ |
5cd4f6fd HQ |
2008 | if (!irqchip_in_kernel(vcpu->kvm)) |
2009 | set_cr8(vcpu, kvm_run->cr8); | |
54810342 | 2010 | |
02c83209 AK |
2011 | if (vcpu->pio.cur_count) { |
2012 | r = complete_pio(vcpu); | |
2013 | if (r) | |
2014 | goto out; | |
2015 | } | |
2016 | ||
2017 | if (vcpu->mmio_needed) { | |
2018 | memcpy(vcpu->mmio_data, kvm_run->mmio.data, 8); | |
2019 | vcpu->mmio_read_completed = 1; | |
2020 | vcpu->mmio_needed = 0; | |
2021 | r = emulate_instruction(vcpu, kvm_run, | |
2022 | vcpu->mmio_fault_cr2, 0); | |
2023 | if (r == EMULATE_DO_MMIO) { | |
2024 | /* | |
2025 | * Read-modify-write. Back to userspace. | |
2026 | */ | |
02c83209 AK |
2027 | r = 0; |
2028 | goto out; | |
46fc1477 | 2029 | } |
6aa8b732 AK |
2030 | } |
2031 | ||
8eb7d334 | 2032 | if (kvm_run->exit_reason == KVM_EXIT_HYPERCALL) { |
b4e63f56 AK |
2033 | kvm_arch_ops->cache_regs(vcpu); |
2034 | vcpu->regs[VCPU_REGS_RAX] = kvm_run->hypercall.ret; | |
2035 | kvm_arch_ops->decache_regs(vcpu); | |
2036 | } | |
2037 | ||
6aa8b732 AK |
2038 | r = kvm_arch_ops->run(vcpu, kvm_run); |
2039 | ||
039576c0 | 2040 | out: |
1961d276 AK |
2041 | if (vcpu->sigset_active) |
2042 | sigprocmask(SIG_SETMASK, &sigsaved, NULL); | |
2043 | ||
6aa8b732 AK |
2044 | vcpu_put(vcpu); |
2045 | return r; | |
2046 | } | |
2047 | ||
bccf2150 AK |
2048 | static int kvm_vcpu_ioctl_get_regs(struct kvm_vcpu *vcpu, |
2049 | struct kvm_regs *regs) | |
6aa8b732 | 2050 | { |
bccf2150 | 2051 | vcpu_load(vcpu); |
6aa8b732 AK |
2052 | |
2053 | kvm_arch_ops->cache_regs(vcpu); | |
2054 | ||
2055 | regs->rax = vcpu->regs[VCPU_REGS_RAX]; | |
2056 | regs->rbx = vcpu->regs[VCPU_REGS_RBX]; | |
2057 | regs->rcx = vcpu->regs[VCPU_REGS_RCX]; | |
2058 | regs->rdx = vcpu->regs[VCPU_REGS_RDX]; | |
2059 | regs->rsi = vcpu->regs[VCPU_REGS_RSI]; | |
2060 | regs->rdi = vcpu->regs[VCPU_REGS_RDI]; | |
2061 | regs->rsp = vcpu->regs[VCPU_REGS_RSP]; | |
2062 | regs->rbp = vcpu->regs[VCPU_REGS_RBP]; | |
05b3e0c2 | 2063 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2064 | regs->r8 = vcpu->regs[VCPU_REGS_R8]; |
2065 | regs->r9 = vcpu->regs[VCPU_REGS_R9]; | |
2066 | regs->r10 = vcpu->regs[VCPU_REGS_R10]; | |
2067 | regs->r11 = vcpu->regs[VCPU_REGS_R11]; | |
2068 | regs->r12 = vcpu->regs[VCPU_REGS_R12]; | |
2069 | regs->r13 = vcpu->regs[VCPU_REGS_R13]; | |
2070 | regs->r14 = vcpu->regs[VCPU_REGS_R14]; | |
2071 | regs->r15 = vcpu->regs[VCPU_REGS_R15]; | |
2072 | #endif | |
2073 | ||
2074 | regs->rip = vcpu->rip; | |
2075 | regs->rflags = kvm_arch_ops->get_rflags(vcpu); | |
2076 | ||
2077 | /* | |
2078 | * Don't leak debug flags in case they were set for guest debugging | |
2079 | */ | |
2080 | if (vcpu->guest_debug.enabled && vcpu->guest_debug.singlestep) | |
2081 | regs->rflags &= ~(X86_EFLAGS_TF | X86_EFLAGS_RF); | |
2082 | ||
2083 | vcpu_put(vcpu); | |
2084 | ||
2085 | return 0; | |
2086 | } | |
2087 | ||
bccf2150 AK |
2088 | static int kvm_vcpu_ioctl_set_regs(struct kvm_vcpu *vcpu, |
2089 | struct kvm_regs *regs) | |
6aa8b732 | 2090 | { |
bccf2150 | 2091 | vcpu_load(vcpu); |
6aa8b732 AK |
2092 | |
2093 | vcpu->regs[VCPU_REGS_RAX] = regs->rax; | |
2094 | vcpu->regs[VCPU_REGS_RBX] = regs->rbx; | |
2095 | vcpu->regs[VCPU_REGS_RCX] = regs->rcx; | |
2096 | vcpu->regs[VCPU_REGS_RDX] = regs->rdx; | |
2097 | vcpu->regs[VCPU_REGS_RSI] = regs->rsi; | |
2098 | vcpu->regs[VCPU_REGS_RDI] = regs->rdi; | |
2099 | vcpu->regs[VCPU_REGS_RSP] = regs->rsp; | |
2100 | vcpu->regs[VCPU_REGS_RBP] = regs->rbp; | |
05b3e0c2 | 2101 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2102 | vcpu->regs[VCPU_REGS_R8] = regs->r8; |
2103 | vcpu->regs[VCPU_REGS_R9] = regs->r9; | |
2104 | vcpu->regs[VCPU_REGS_R10] = regs->r10; | |
2105 | vcpu->regs[VCPU_REGS_R11] = regs->r11; | |
2106 | vcpu->regs[VCPU_REGS_R12] = regs->r12; | |
2107 | vcpu->regs[VCPU_REGS_R13] = regs->r13; | |
2108 | vcpu->regs[VCPU_REGS_R14] = regs->r14; | |
2109 | vcpu->regs[VCPU_REGS_R15] = regs->r15; | |
2110 | #endif | |
2111 | ||
2112 | vcpu->rip = regs->rip; | |
2113 | kvm_arch_ops->set_rflags(vcpu, regs->rflags); | |
2114 | ||
2115 | kvm_arch_ops->decache_regs(vcpu); | |
2116 | ||
2117 | vcpu_put(vcpu); | |
2118 | ||
2119 | return 0; | |
2120 | } | |
2121 | ||
2122 | static void get_segment(struct kvm_vcpu *vcpu, | |
2123 | struct kvm_segment *var, int seg) | |
2124 | { | |
2125 | return kvm_arch_ops->get_segment(vcpu, var, seg); | |
2126 | } | |
2127 | ||
bccf2150 AK |
2128 | static int kvm_vcpu_ioctl_get_sregs(struct kvm_vcpu *vcpu, |
2129 | struct kvm_sregs *sregs) | |
6aa8b732 | 2130 | { |
6aa8b732 | 2131 | struct descriptor_table dt; |
2a8067f1 | 2132 | int pending_vec; |
6aa8b732 | 2133 | |
bccf2150 | 2134 | vcpu_load(vcpu); |
6aa8b732 AK |
2135 | |
2136 | get_segment(vcpu, &sregs->cs, VCPU_SREG_CS); | |
2137 | get_segment(vcpu, &sregs->ds, VCPU_SREG_DS); | |
2138 | get_segment(vcpu, &sregs->es, VCPU_SREG_ES); | |
2139 | get_segment(vcpu, &sregs->fs, VCPU_SREG_FS); | |
2140 | get_segment(vcpu, &sregs->gs, VCPU_SREG_GS); | |
2141 | get_segment(vcpu, &sregs->ss, VCPU_SREG_SS); | |
2142 | ||
2143 | get_segment(vcpu, &sregs->tr, VCPU_SREG_TR); | |
2144 | get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); | |
2145 | ||
2146 | kvm_arch_ops->get_idt(vcpu, &dt); | |
2147 | sregs->idt.limit = dt.limit; | |
2148 | sregs->idt.base = dt.base; | |
2149 | kvm_arch_ops->get_gdt(vcpu, &dt); | |
2150 | sregs->gdt.limit = dt.limit; | |
2151 | sregs->gdt.base = dt.base; | |
2152 | ||
25c4c276 | 2153 | kvm_arch_ops->decache_cr4_guest_bits(vcpu); |
6aa8b732 AK |
2154 | sregs->cr0 = vcpu->cr0; |
2155 | sregs->cr2 = vcpu->cr2; | |
2156 | sregs->cr3 = vcpu->cr3; | |
2157 | sregs->cr4 = vcpu->cr4; | |
7017fc3d | 2158 | sregs->cr8 = get_cr8(vcpu); |
6aa8b732 | 2159 | sregs->efer = vcpu->shadow_efer; |
7017fc3d | 2160 | sregs->apic_base = kvm_get_apic_base(vcpu); |
6aa8b732 | 2161 | |
2a8067f1 | 2162 | if (irqchip_in_kernel(vcpu->kvm)) { |
c52fb35a HQ |
2163 | memset(sregs->interrupt_bitmap, 0, |
2164 | sizeof sregs->interrupt_bitmap); | |
2a8067f1 ED |
2165 | pending_vec = kvm_arch_ops->get_irq(vcpu); |
2166 | if (pending_vec >= 0) | |
2167 | set_bit(pending_vec, (unsigned long *)sregs->interrupt_bitmap); | |
2168 | } else | |
c52fb35a HQ |
2169 | memcpy(sregs->interrupt_bitmap, vcpu->irq_pending, |
2170 | sizeof sregs->interrupt_bitmap); | |
6aa8b732 AK |
2171 | |
2172 | vcpu_put(vcpu); | |
2173 | ||
2174 | return 0; | |
2175 | } | |
2176 | ||
2177 | static void set_segment(struct kvm_vcpu *vcpu, | |
2178 | struct kvm_segment *var, int seg) | |
2179 | { | |
2180 | return kvm_arch_ops->set_segment(vcpu, var, seg); | |
2181 | } | |
2182 | ||
bccf2150 AK |
2183 | static int kvm_vcpu_ioctl_set_sregs(struct kvm_vcpu *vcpu, |
2184 | struct kvm_sregs *sregs) | |
6aa8b732 | 2185 | { |
6aa8b732 | 2186 | int mmu_reset_needed = 0; |
2a8067f1 | 2187 | int i, pending_vec, max_bits; |
6aa8b732 AK |
2188 | struct descriptor_table dt; |
2189 | ||
bccf2150 | 2190 | vcpu_load(vcpu); |
6aa8b732 | 2191 | |
6aa8b732 AK |
2192 | dt.limit = sregs->idt.limit; |
2193 | dt.base = sregs->idt.base; | |
2194 | kvm_arch_ops->set_idt(vcpu, &dt); | |
2195 | dt.limit = sregs->gdt.limit; | |
2196 | dt.base = sregs->gdt.base; | |
2197 | kvm_arch_ops->set_gdt(vcpu, &dt); | |
2198 | ||
2199 | vcpu->cr2 = sregs->cr2; | |
2200 | mmu_reset_needed |= vcpu->cr3 != sregs->cr3; | |
2201 | vcpu->cr3 = sregs->cr3; | |
2202 | ||
7017fc3d | 2203 | set_cr8(vcpu, sregs->cr8); |
6aa8b732 AK |
2204 | |
2205 | mmu_reset_needed |= vcpu->shadow_efer != sregs->efer; | |
05b3e0c2 | 2206 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2207 | kvm_arch_ops->set_efer(vcpu, sregs->efer); |
2208 | #endif | |
7017fc3d | 2209 | kvm_set_apic_base(vcpu, sregs->apic_base); |
6aa8b732 | 2210 | |
25c4c276 | 2211 | kvm_arch_ops->decache_cr4_guest_bits(vcpu); |
399badf3 | 2212 | |
6aa8b732 | 2213 | mmu_reset_needed |= vcpu->cr0 != sregs->cr0; |
f6528b03 | 2214 | kvm_arch_ops->set_cr0(vcpu, sregs->cr0); |
6aa8b732 AK |
2215 | |
2216 | mmu_reset_needed |= vcpu->cr4 != sregs->cr4; | |
2217 | kvm_arch_ops->set_cr4(vcpu, sregs->cr4); | |
1b0973bd AK |
2218 | if (!is_long_mode(vcpu) && is_pae(vcpu)) |
2219 | load_pdptrs(vcpu, vcpu->cr3); | |
6aa8b732 AK |
2220 | |
2221 | if (mmu_reset_needed) | |
2222 | kvm_mmu_reset_context(vcpu); | |
2223 | ||
c52fb35a HQ |
2224 | if (!irqchip_in_kernel(vcpu->kvm)) { |
2225 | memcpy(vcpu->irq_pending, sregs->interrupt_bitmap, | |
2226 | sizeof vcpu->irq_pending); | |
2227 | vcpu->irq_summary = 0; | |
2228 | for (i = 0; i < ARRAY_SIZE(vcpu->irq_pending); ++i) | |
2229 | if (vcpu->irq_pending[i]) | |
2230 | __set_bit(i, &vcpu->irq_summary); | |
2a8067f1 ED |
2231 | } else { |
2232 | max_bits = (sizeof sregs->interrupt_bitmap) << 3; | |
2233 | pending_vec = find_first_bit( | |
2234 | (const unsigned long *)sregs->interrupt_bitmap, | |
2235 | max_bits); | |
2236 | /* Only pending external irq is handled here */ | |
2237 | if (pending_vec < max_bits) { | |
2238 | kvm_arch_ops->set_irq(vcpu, pending_vec); | |
2239 | printk("Set back pending irq %d\n", pending_vec); | |
2240 | } | |
c52fb35a | 2241 | } |
6aa8b732 | 2242 | |
024aa1c0 AK |
2243 | set_segment(vcpu, &sregs->cs, VCPU_SREG_CS); |
2244 | set_segment(vcpu, &sregs->ds, VCPU_SREG_DS); | |
2245 | set_segment(vcpu, &sregs->es, VCPU_SREG_ES); | |
2246 | set_segment(vcpu, &sregs->fs, VCPU_SREG_FS); | |
2247 | set_segment(vcpu, &sregs->gs, VCPU_SREG_GS); | |
2248 | set_segment(vcpu, &sregs->ss, VCPU_SREG_SS); | |
2249 | ||
2250 | set_segment(vcpu, &sregs->tr, VCPU_SREG_TR); | |
2251 | set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR); | |
2252 | ||
6aa8b732 AK |
2253 | vcpu_put(vcpu); |
2254 | ||
2255 | return 0; | |
2256 | } | |
2257 | ||
2258 | /* | |
2259 | * List of msr numbers which we expose to userspace through KVM_GET_MSRS | |
2260 | * and KVM_SET_MSRS, and KVM_GET_MSR_INDEX_LIST. | |
bf591b24 MR |
2261 | * |
2262 | * This list is modified at module load time to reflect the | |
2263 | * capabilities of the host cpu. | |
6aa8b732 AK |
2264 | */ |
2265 | static u32 msrs_to_save[] = { | |
2266 | MSR_IA32_SYSENTER_CS, MSR_IA32_SYSENTER_ESP, MSR_IA32_SYSENTER_EIP, | |
2267 | MSR_K6_STAR, | |
05b3e0c2 | 2268 | #ifdef CONFIG_X86_64 |
6aa8b732 AK |
2269 | MSR_CSTAR, MSR_KERNEL_GS_BASE, MSR_SYSCALL_MASK, MSR_LSTAR, |
2270 | #endif | |
2271 | MSR_IA32_TIME_STAMP_COUNTER, | |
2272 | }; | |
2273 | ||
bf591b24 MR |
2274 | static unsigned num_msrs_to_save; |
2275 | ||
6f00e68f AK |
2276 | static u32 emulated_msrs[] = { |
2277 | MSR_IA32_MISC_ENABLE, | |
2278 | }; | |
2279 | ||
bf591b24 MR |
2280 | static __init void kvm_init_msr_list(void) |
2281 | { | |
2282 | u32 dummy[2]; | |
2283 | unsigned i, j; | |
2284 | ||
2285 | for (i = j = 0; i < ARRAY_SIZE(msrs_to_save); i++) { | |
2286 | if (rdmsr_safe(msrs_to_save[i], &dummy[0], &dummy[1]) < 0) | |
2287 | continue; | |
2288 | if (j < i) | |
2289 | msrs_to_save[j] = msrs_to_save[i]; | |
2290 | j++; | |
2291 | } | |
2292 | num_msrs_to_save = j; | |
2293 | } | |
6aa8b732 AK |
2294 | |
2295 | /* | |
2296 | * Adapt set_msr() to msr_io()'s calling convention | |
2297 | */ | |
2298 | static int do_set_msr(struct kvm_vcpu *vcpu, unsigned index, u64 *data) | |
2299 | { | |
35f3f286 | 2300 | return kvm_set_msr(vcpu, index, *data); |
6aa8b732 AK |
2301 | } |
2302 | ||
2303 | /* | |
2304 | * Read or write a bunch of msrs. All parameters are kernel addresses. | |
2305 | * | |
2306 | * @return number of msrs set successfully. | |
2307 | */ | |
bccf2150 | 2308 | static int __msr_io(struct kvm_vcpu *vcpu, struct kvm_msrs *msrs, |
6aa8b732 AK |
2309 | struct kvm_msr_entry *entries, |
2310 | int (*do_msr)(struct kvm_vcpu *vcpu, | |
2311 | unsigned index, u64 *data)) | |
2312 | { | |
6aa8b732 AK |
2313 | int i; |
2314 | ||
bccf2150 | 2315 | vcpu_load(vcpu); |
6aa8b732 AK |
2316 | |
2317 | for (i = 0; i < msrs->nmsrs; ++i) | |
2318 | if (do_msr(vcpu, entries[i].index, &entries[i].data)) | |
2319 | break; | |
2320 | ||
2321 | vcpu_put(vcpu); | |
2322 | ||
2323 | return i; | |
2324 | } | |
2325 | ||
2326 | /* | |
2327 | * Read or write a bunch of msrs. Parameters are user addresses. | |
2328 | * | |
2329 | * @return number of msrs set successfully. | |
2330 | */ | |
bccf2150 | 2331 | static int msr_io(struct kvm_vcpu *vcpu, struct kvm_msrs __user *user_msrs, |
6aa8b732 AK |
2332 | int (*do_msr)(struct kvm_vcpu *vcpu, |
2333 | unsigned index, u64 *data), | |
2334 | int writeback) | |
2335 | { | |
2336 | struct kvm_msrs msrs; | |
2337 | struct kvm_msr_entry *entries; | |
2338 | int r, n; | |
2339 | unsigned size; | |
2340 | ||
2341 | r = -EFAULT; | |
2342 | if (copy_from_user(&msrs, user_msrs, sizeof msrs)) | |
2343 | goto out; | |
2344 | ||
2345 | r = -E2BIG; | |
2346 | if (msrs.nmsrs >= MAX_IO_MSRS) | |
2347 | goto out; | |
2348 | ||
2349 | r = -ENOMEM; | |
2350 | size = sizeof(struct kvm_msr_entry) * msrs.nmsrs; | |
2351 | entries = vmalloc(size); | |
2352 | if (!entries) | |
2353 | goto out; | |
2354 | ||
2355 | r = -EFAULT; | |
2356 | if (copy_from_user(entries, user_msrs->entries, size)) | |
2357 | goto out_free; | |
2358 | ||
bccf2150 | 2359 | r = n = __msr_io(vcpu, &msrs, entries, do_msr); |
6aa8b732 AK |
2360 | if (r < 0) |
2361 | goto out_free; | |
2362 | ||
2363 | r = -EFAULT; | |
2364 | if (writeback && copy_to_user(user_msrs->entries, entries, size)) | |
2365 | goto out_free; | |
2366 | ||
2367 | r = n; | |
2368 | ||
2369 | out_free: | |
2370 | vfree(entries); | |
2371 | out: | |
2372 | return r; | |
2373 | } | |
2374 | ||
2375 | /* | |
2376 | * Translate a guest virtual address to a guest physical address. | |
2377 | */ | |
bccf2150 AK |
2378 | static int kvm_vcpu_ioctl_translate(struct kvm_vcpu *vcpu, |
2379 | struct kvm_translation *tr) | |
6aa8b732 AK |
2380 | { |
2381 | unsigned long vaddr = tr->linear_address; | |
6aa8b732 AK |
2382 | gpa_t gpa; |
2383 | ||
bccf2150 | 2384 | vcpu_load(vcpu); |
11ec2804 | 2385 | mutex_lock(&vcpu->kvm->lock); |
6aa8b732 AK |
2386 | gpa = vcpu->mmu.gva_to_gpa(vcpu, vaddr); |
2387 | tr->physical_address = gpa; | |
2388 | tr->valid = gpa != UNMAPPED_GVA; | |
2389 | tr->writeable = 1; | |
2390 | tr->usermode = 0; | |
11ec2804 | 2391 | mutex_unlock(&vcpu->kvm->lock); |
6aa8b732 AK |
2392 | vcpu_put(vcpu); |
2393 | ||
2394 | return 0; | |
2395 | } | |
2396 | ||
bccf2150 AK |
2397 | static int kvm_vcpu_ioctl_interrupt(struct kvm_vcpu *vcpu, |
2398 | struct kvm_interrupt *irq) | |
6aa8b732 | 2399 | { |
6aa8b732 AK |
2400 | if (irq->irq < 0 || irq->irq >= 256) |
2401 | return -EINVAL; | |
97222cc8 ED |
2402 | if (irqchip_in_kernel(vcpu->kvm)) |
2403 | return -ENXIO; | |
bccf2150 | 2404 | vcpu_load(vcpu); |
6aa8b732 AK |
2405 | |
2406 | set_bit(irq->irq, vcpu->irq_pending); | |
2407 | set_bit(irq->irq / BITS_PER_LONG, &vcpu->irq_summary); | |
2408 | ||
2409 | vcpu_put(vcpu); | |
2410 | ||
2411 | return 0; | |
2412 | } | |
2413 | ||
bccf2150 AK |
2414 | static int kvm_vcpu_ioctl_debug_guest(struct kvm_vcpu *vcpu, |
2415 | struct kvm_debug_guest *dbg) | |
6aa8b732 | 2416 | { |
6aa8b732 AK |
2417 | int r; |
2418 | ||
bccf2150 | 2419 | vcpu_load(vcpu); |
6aa8b732 AK |
2420 | |
2421 | r = kvm_arch_ops->set_guest_debug(vcpu, dbg); | |
2422 | ||
2423 | vcpu_put(vcpu); | |
2424 | ||
2425 | return r; | |
2426 | } | |
2427 | ||
9a2bb7f4 AK |
2428 | static struct page *kvm_vcpu_nopage(struct vm_area_struct *vma, |
2429 | unsigned long address, | |
2430 | int *type) | |
2431 | { | |
2432 | struct kvm_vcpu *vcpu = vma->vm_file->private_data; | |
2433 | unsigned long pgoff; | |
2434 | struct page *page; | |
2435 | ||
9a2bb7f4 | 2436 | pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
039576c0 AK |
2437 | if (pgoff == 0) |
2438 | page = virt_to_page(vcpu->run); | |
2439 | else if (pgoff == KVM_PIO_PAGE_OFFSET) | |
2440 | page = virt_to_page(vcpu->pio_data); | |
2441 | else | |
9a2bb7f4 | 2442 | return NOPAGE_SIGBUS; |
9a2bb7f4 | 2443 | get_page(page); |
cd0d9137 NAQ |
2444 | if (type != NULL) |
2445 | *type = VM_FAULT_MINOR; | |
2446 | ||
9a2bb7f4 AK |
2447 | return page; |
2448 | } | |
2449 | ||
2450 | static struct vm_operations_struct kvm_vcpu_vm_ops = { | |
2451 | .nopage = kvm_vcpu_nopage, | |
2452 | }; | |
2453 | ||
2454 | static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma) | |
2455 | { | |
2456 | vma->vm_ops = &kvm_vcpu_vm_ops; | |
2457 | return 0; | |
2458 | } | |
2459 | ||
bccf2150 AK |
2460 | static int kvm_vcpu_release(struct inode *inode, struct file *filp) |
2461 | { | |
2462 | struct kvm_vcpu *vcpu = filp->private_data; | |
2463 | ||
2464 | fput(vcpu->kvm->filp); | |
2465 | return 0; | |
2466 | } | |
2467 | ||
2468 | static struct file_operations kvm_vcpu_fops = { | |
2469 | .release = kvm_vcpu_release, | |
2470 | .unlocked_ioctl = kvm_vcpu_ioctl, | |
2471 | .compat_ioctl = kvm_vcpu_ioctl, | |
9a2bb7f4 | 2472 | .mmap = kvm_vcpu_mmap, |
bccf2150 AK |
2473 | }; |
2474 | ||
2475 | /* | |
2476 | * Allocates an inode for the vcpu. | |
2477 | */ | |
2478 | static int create_vcpu_fd(struct kvm_vcpu *vcpu) | |
2479 | { | |
2480 | int fd, r; | |
2481 | struct inode *inode; | |
2482 | struct file *file; | |
2483 | ||
d6d28168 AK |
2484 | r = anon_inode_getfd(&fd, &inode, &file, |
2485 | "kvm-vcpu", &kvm_vcpu_fops, vcpu); | |
2486 | if (r) | |
2487 | return r; | |
bccf2150 | 2488 | atomic_inc(&vcpu->kvm->filp->f_count); |
bccf2150 | 2489 | return fd; |
bccf2150 AK |
2490 | } |
2491 | ||
c5ea7660 AK |
2492 | /* |
2493 | * Creates some virtual cpus. Good luck creating more than one. | |
2494 | */ | |
2495 | static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n) | |
2496 | { | |
2497 | int r; | |
2498 | struct kvm_vcpu *vcpu; | |
2499 | ||
c5ea7660 | 2500 | if (!valid_vcpu(n)) |
fb3f0f51 | 2501 | return -EINVAL; |
c5ea7660 | 2502 | |
fb3f0f51 RR |
2503 | vcpu = kvm_arch_ops->vcpu_create(kvm, n); |
2504 | if (IS_ERR(vcpu)) | |
2505 | return PTR_ERR(vcpu); | |
c5ea7660 | 2506 | |
15ad7146 AK |
2507 | preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops); |
2508 | ||
b114b080 RR |
2509 | /* We do fxsave: this must be aligned. */ |
2510 | BUG_ON((unsigned long)&vcpu->host_fx_image & 0xF); | |
2511 | ||
fb3f0f51 | 2512 | vcpu_load(vcpu); |
c5ea7660 | 2513 | r = kvm_mmu_setup(vcpu); |
c5ea7660 | 2514 | vcpu_put(vcpu); |
c5ea7660 | 2515 | if (r < 0) |
fb3f0f51 RR |
2516 | goto free_vcpu; |
2517 | ||
11ec2804 | 2518 | mutex_lock(&kvm->lock); |
fb3f0f51 RR |
2519 | if (kvm->vcpus[n]) { |
2520 | r = -EEXIST; | |
11ec2804 | 2521 | mutex_unlock(&kvm->lock); |
fb3f0f51 RR |
2522 | goto mmu_unload; |
2523 | } | |
2524 | kvm->vcpus[n] = vcpu; | |
11ec2804 | 2525 | mutex_unlock(&kvm->lock); |
c5ea7660 | 2526 | |
fb3f0f51 | 2527 | /* Now it's all set up, let userspace reach it */ |
bccf2150 AK |
2528 | r = create_vcpu_fd(vcpu); |
2529 | if (r < 0) | |
fb3f0f51 RR |
2530 | goto unlink; |
2531 | return r; | |
39c3b86e | 2532 | |
fb3f0f51 | 2533 | unlink: |
11ec2804 | 2534 | mutex_lock(&kvm->lock); |
fb3f0f51 | 2535 | kvm->vcpus[n] = NULL; |
11ec2804 | 2536 | mutex_unlock(&kvm->lock); |
a2fa3e9f | 2537 | |
fb3f0f51 RR |
2538 | mmu_unload: |
2539 | vcpu_load(vcpu); | |
2540 | kvm_mmu_unload(vcpu); | |
2541 | vcpu_put(vcpu); | |
c5ea7660 | 2542 | |
fb3f0f51 RR |
2543 | free_vcpu: |
2544 | kvm_arch_ops->vcpu_free(vcpu); | |
c5ea7660 AK |
2545 | return r; |
2546 | } | |
2547 | ||
2cc51560 ED |
2548 | static void cpuid_fix_nx_cap(struct kvm_vcpu *vcpu) |
2549 | { | |
2550 | u64 efer; | |
2551 | int i; | |
2552 | struct kvm_cpuid_entry *e, *entry; | |
2553 | ||
2554 | rdmsrl(MSR_EFER, efer); | |
2555 | entry = NULL; | |
2556 | for (i = 0; i < vcpu->cpuid_nent; ++i) { | |
2557 | e = &vcpu->cpuid_entries[i]; | |
2558 | if (e->function == 0x80000001) { | |
2559 | entry = e; | |
2560 | break; | |
2561 | } | |
2562 | } | |
4c981b43 | 2563 | if (entry && (entry->edx & (1 << 20)) && !(efer & EFER_NX)) { |
2cc51560 | 2564 | entry->edx &= ~(1 << 20); |
4c981b43 | 2565 | printk(KERN_INFO "kvm: guest NX capability removed\n"); |
2cc51560 ED |
2566 | } |
2567 | } | |
2568 | ||
06465c5a AK |
2569 | static int kvm_vcpu_ioctl_set_cpuid(struct kvm_vcpu *vcpu, |
2570 | struct kvm_cpuid *cpuid, | |
2571 | struct kvm_cpuid_entry __user *entries) | |
2572 | { | |
2573 | int r; | |
2574 | ||
2575 | r = -E2BIG; | |
2576 | if (cpuid->nent > KVM_MAX_CPUID_ENTRIES) | |
2577 | goto out; | |
2578 | r = -EFAULT; | |
2579 | if (copy_from_user(&vcpu->cpuid_entries, entries, | |
2580 | cpuid->nent * sizeof(struct kvm_cpuid_entry))) | |
2581 | goto out; | |
2582 | vcpu->cpuid_nent = cpuid->nent; | |
2cc51560 | 2583 | cpuid_fix_nx_cap(vcpu); |
06465c5a AK |
2584 | return 0; |
2585 | ||
2586 | out: | |
2587 | return r; | |
2588 | } | |
2589 | ||
1961d276 AK |
2590 | static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset) |
2591 | { | |
2592 | if (sigset) { | |
2593 | sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP)); | |
2594 | vcpu->sigset_active = 1; | |
2595 | vcpu->sigset = *sigset; | |
2596 | } else | |
2597 | vcpu->sigset_active = 0; | |
2598 | return 0; | |
2599 | } | |
2600 | ||
b8836737 AK |
2601 | /* |
2602 | * fxsave fpu state. Taken from x86_64/processor.h. To be killed when | |
2603 | * we have asm/x86/processor.h | |
2604 | */ | |
2605 | struct fxsave { | |
2606 | u16 cwd; | |
2607 | u16 swd; | |
2608 | u16 twd; | |
2609 | u16 fop; | |
2610 | u64 rip; | |
2611 | u64 rdp; | |
2612 | u32 mxcsr; | |
2613 | u32 mxcsr_mask; | |
2614 | u32 st_space[32]; /* 8*16 bytes for each FP-reg = 128 bytes */ | |
2615 | #ifdef CONFIG_X86_64 | |
2616 | u32 xmm_space[64]; /* 16*16 bytes for each XMM-reg = 256 bytes */ | |
2617 | #else | |
2618 | u32 xmm_space[32]; /* 8*16 bytes for each XMM-reg = 128 bytes */ | |
2619 | #endif | |
2620 | }; | |
2621 | ||
2622 | static int kvm_vcpu_ioctl_get_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
2623 | { | |
b114b080 | 2624 | struct fxsave *fxsave = (struct fxsave *)&vcpu->guest_fx_image; |
b8836737 AK |
2625 | |
2626 | vcpu_load(vcpu); | |
2627 | ||
2628 | memcpy(fpu->fpr, fxsave->st_space, 128); | |
2629 | fpu->fcw = fxsave->cwd; | |
2630 | fpu->fsw = fxsave->swd; | |
2631 | fpu->ftwx = fxsave->twd; | |
2632 | fpu->last_opcode = fxsave->fop; | |
2633 | fpu->last_ip = fxsave->rip; | |
2634 | fpu->last_dp = fxsave->rdp; | |
2635 | memcpy(fpu->xmm, fxsave->xmm_space, sizeof fxsave->xmm_space); | |
2636 | ||
2637 | vcpu_put(vcpu); | |
2638 | ||
2639 | return 0; | |
2640 | } | |
2641 | ||
2642 | static int kvm_vcpu_ioctl_set_fpu(struct kvm_vcpu *vcpu, struct kvm_fpu *fpu) | |
2643 | { | |
b114b080 | 2644 | struct fxsave *fxsave = (struct fxsave *)&vcpu->guest_fx_image; |
b8836737 AK |
2645 | |
2646 | vcpu_load(vcpu); | |
2647 | ||
2648 | memcpy(fxsave->st_space, fpu->fpr, 128); | |
2649 | fxsave->cwd = fpu->fcw; | |
2650 | fxsave->swd = fpu->fsw; | |
2651 | fxsave->twd = fpu->ftwx; | |
2652 | fxsave->fop = fpu->last_opcode; | |
2653 | fxsave->rip = fpu->last_ip; | |
2654 | fxsave->rdp = fpu->last_dp; | |
2655 | memcpy(fxsave->xmm_space, fpu->xmm, sizeof fxsave->xmm_space); | |
2656 | ||
2657 | vcpu_put(vcpu); | |
2658 | ||
2659 | return 0; | |
2660 | } | |
2661 | ||
96ad2cc6 ED |
2662 | static int kvm_vcpu_ioctl_get_lapic(struct kvm_vcpu *vcpu, |
2663 | struct kvm_lapic_state *s) | |
2664 | { | |
2665 | vcpu_load(vcpu); | |
2666 | memcpy(s->regs, vcpu->apic->regs, sizeof *s); | |
2667 | vcpu_put(vcpu); | |
2668 | ||
2669 | return 0; | |
2670 | } | |
2671 | ||
2672 | static int kvm_vcpu_ioctl_set_lapic(struct kvm_vcpu *vcpu, | |
2673 | struct kvm_lapic_state *s) | |
2674 | { | |
2675 | vcpu_load(vcpu); | |
2676 | memcpy(vcpu->apic->regs, s->regs, sizeof *s); | |
2677 | kvm_apic_post_state_restore(vcpu); | |
2678 | vcpu_put(vcpu); | |
2679 | ||
2680 | return 0; | |
2681 | } | |
2682 | ||
bccf2150 AK |
2683 | static long kvm_vcpu_ioctl(struct file *filp, |
2684 | unsigned int ioctl, unsigned long arg) | |
6aa8b732 | 2685 | { |
bccf2150 | 2686 | struct kvm_vcpu *vcpu = filp->private_data; |
2f366987 | 2687 | void __user *argp = (void __user *)arg; |
6aa8b732 AK |
2688 | int r = -EINVAL; |
2689 | ||
2690 | switch (ioctl) { | |
9a2bb7f4 | 2691 | case KVM_RUN: |
f0fe5108 AK |
2692 | r = -EINVAL; |
2693 | if (arg) | |
2694 | goto out; | |
9a2bb7f4 | 2695 | r = kvm_vcpu_ioctl_run(vcpu, vcpu->run); |
6aa8b732 | 2696 | break; |
6aa8b732 AK |
2697 | case KVM_GET_REGS: { |
2698 | struct kvm_regs kvm_regs; | |
2699 | ||
bccf2150 AK |
2700 | memset(&kvm_regs, 0, sizeof kvm_regs); |
2701 | r = kvm_vcpu_ioctl_get_regs(vcpu, &kvm_regs); | |
6aa8b732 AK |
2702 | if (r) |
2703 | goto out; | |
2704 | r = -EFAULT; | |
2f366987 | 2705 | if (copy_to_user(argp, &kvm_regs, sizeof kvm_regs)) |
6aa8b732 AK |
2706 | goto out; |
2707 | r = 0; | |
2708 | break; | |
2709 | } | |
2710 | case KVM_SET_REGS: { | |
2711 | struct kvm_regs kvm_regs; | |
2712 | ||
2713 | r = -EFAULT; | |
2f366987 | 2714 | if (copy_from_user(&kvm_regs, argp, sizeof kvm_regs)) |
6aa8b732 | 2715 | goto out; |
bccf2150 | 2716 | r = kvm_vcpu_ioctl_set_regs(vcpu, &kvm_regs); |
6aa8b732 AK |
2717 | if (r) |
2718 | goto out; | |
2719 | r = 0; | |
2720 | break; | |
2721 | } | |
2722 | case KVM_GET_SREGS: { | |
2723 | struct kvm_sregs kvm_sregs; | |
2724 | ||
bccf2150 AK |
2725 | memset(&kvm_sregs, 0, sizeof kvm_sregs); |
2726 | r = kvm_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs); | |
6aa8b732 AK |
2727 | if (r) |
2728 | goto out; | |
2729 | r = -EFAULT; | |
2f366987 | 2730 | if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs)) |
6aa8b732 AK |
2731 | goto out; |
2732 | r = 0; | |
2733 | break; | |
2734 | } | |
2735 | case KVM_SET_SREGS: { | |
2736 | struct kvm_sregs kvm_sregs; | |
2737 | ||
2738 | r = -EFAULT; | |
2f366987 | 2739 | if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs)) |
6aa8b732 | 2740 | goto out; |
bccf2150 | 2741 | r = kvm_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs); |
6aa8b732 AK |
2742 | if (r) |
2743 | goto out; | |
2744 | r = 0; | |
2745 | break; | |
2746 | } | |
2747 | case KVM_TRANSLATE: { | |
2748 | struct kvm_translation tr; | |
2749 | ||
2750 | r = -EFAULT; | |
2f366987 | 2751 | if (copy_from_user(&tr, argp, sizeof tr)) |
6aa8b732 | 2752 | goto out; |
bccf2150 | 2753 | r = kvm_vcpu_ioctl_translate(vcpu, &tr); |
6aa8b732 AK |
2754 | if (r) |
2755 | goto out; | |
2756 | r = -EFAULT; | |
2f366987 | 2757 | if (copy_to_user(argp, &tr, sizeof tr)) |
6aa8b732 AK |
2758 | goto out; |
2759 | r = 0; | |
2760 | break; | |
2761 | } | |
2762 | case KVM_INTERRUPT: { | |
2763 | struct kvm_interrupt irq; | |
2764 | ||
2765 | r = -EFAULT; | |
2f366987 | 2766 | if (copy_from_user(&irq, argp, sizeof irq)) |
6aa8b732 | 2767 | goto out; |
bccf2150 | 2768 | r = kvm_vcpu_ioctl_interrupt(vcpu, &irq); |
6aa8b732 AK |
2769 | if (r) |
2770 | goto out; | |
2771 | r = 0; | |
2772 | break; | |
2773 | } | |
2774 | case KVM_DEBUG_GUEST: { | |
2775 | struct kvm_debug_guest dbg; | |
2776 | ||
2777 | r = -EFAULT; | |
2f366987 | 2778 | if (copy_from_user(&dbg, argp, sizeof dbg)) |
6aa8b732 | 2779 | goto out; |
bccf2150 | 2780 | r = kvm_vcpu_ioctl_debug_guest(vcpu, &dbg); |
6aa8b732 AK |
2781 | if (r) |
2782 | goto out; | |
2783 | r = 0; | |
2784 | break; | |
2785 | } | |
bccf2150 | 2786 | case KVM_GET_MSRS: |
35f3f286 | 2787 | r = msr_io(vcpu, argp, kvm_get_msr, 1); |
bccf2150 AK |
2788 | break; |
2789 | case KVM_SET_MSRS: | |
2790 | r = msr_io(vcpu, argp, do_set_msr, 0); | |
2791 | break; | |
06465c5a AK |
2792 | case KVM_SET_CPUID: { |
2793 | struct kvm_cpuid __user *cpuid_arg = argp; | |
2794 | struct kvm_cpuid cpuid; | |
2795 | ||
2796 | r = -EFAULT; | |
2797 | if (copy_from_user(&cpuid, cpuid_arg, sizeof cpuid)) | |
2798 | goto out; | |
2799 | r = kvm_vcpu_ioctl_set_cpuid(vcpu, &cpuid, cpuid_arg->entries); | |
2800 | if (r) | |
2801 | goto out; | |
2802 | break; | |
2803 | } | |
1961d276 AK |
2804 | case KVM_SET_SIGNAL_MASK: { |
2805 | struct kvm_signal_mask __user *sigmask_arg = argp; | |
2806 | struct kvm_signal_mask kvm_sigmask; | |
2807 | sigset_t sigset, *p; | |
2808 | ||
2809 | p = NULL; | |
2810 | if (argp) { | |
2811 | r = -EFAULT; | |
2812 | if (copy_from_user(&kvm_sigmask, argp, | |
2813 | sizeof kvm_sigmask)) | |
2814 | goto out; | |
2815 | r = -EINVAL; | |
2816 | if (kvm_sigmask.len != sizeof sigset) | |
2817 | goto out; | |
2818 | r = -EFAULT; | |
2819 | if (copy_from_user(&sigset, sigmask_arg->sigset, | |
2820 | sizeof sigset)) | |
2821 | goto out; | |
2822 | p = &sigset; | |
2823 | } | |
2824 | r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset); | |
2825 | break; | |
2826 | } | |
b8836737 AK |
2827 | case KVM_GET_FPU: { |
2828 | struct kvm_fpu fpu; | |
2829 | ||
2830 | memset(&fpu, 0, sizeof fpu); | |
2831 | r = kvm_vcpu_ioctl_get_fpu(vcpu, &fpu); | |
2832 | if (r) | |
2833 | goto out; | |
2834 | r = -EFAULT; | |
2835 | if (copy_to_user(argp, &fpu, sizeof fpu)) | |
2836 | goto out; | |
2837 | r = 0; | |
2838 | break; | |
2839 | } | |
2840 | case KVM_SET_FPU: { | |
2841 | struct kvm_fpu fpu; | |
2842 | ||
2843 | r = -EFAULT; | |
2844 | if (copy_from_user(&fpu, argp, sizeof fpu)) | |
2845 | goto out; | |
2846 | r = kvm_vcpu_ioctl_set_fpu(vcpu, &fpu); | |
2847 | if (r) | |
2848 | goto out; | |
2849 | r = 0; | |
2850 | break; | |
2851 | } | |
96ad2cc6 ED |
2852 | case KVM_GET_LAPIC: { |
2853 | struct kvm_lapic_state lapic; | |
2854 | ||
2855 | memset(&lapic, 0, sizeof lapic); | |
2856 | r = kvm_vcpu_ioctl_get_lapic(vcpu, &lapic); | |
2857 | if (r) | |
2858 | goto out; | |
2859 | r = -EFAULT; | |
2860 | if (copy_to_user(argp, &lapic, sizeof lapic)) | |
2861 | goto out; | |
2862 | r = 0; | |
2863 | break; | |
2864 | } | |
2865 | case KVM_SET_LAPIC: { | |
2866 | struct kvm_lapic_state lapic; | |
2867 | ||
2868 | r = -EFAULT; | |
2869 | if (copy_from_user(&lapic, argp, sizeof lapic)) | |
2870 | goto out; | |
2871 | r = kvm_vcpu_ioctl_set_lapic(vcpu, &lapic);; | |
2872 | if (r) | |
2873 | goto out; | |
2874 | r = 0; | |
2875 | break; | |
2876 | } | |
bccf2150 AK |
2877 | default: |
2878 | ; | |
2879 | } | |
2880 | out: | |
2881 | return r; | |
2882 | } | |
2883 | ||
2884 | static long kvm_vm_ioctl(struct file *filp, | |
2885 | unsigned int ioctl, unsigned long arg) | |
2886 | { | |
2887 | struct kvm *kvm = filp->private_data; | |
2888 | void __user *argp = (void __user *)arg; | |
2889 | int r = -EINVAL; | |
2890 | ||
2891 | switch (ioctl) { | |
2892 | case KVM_CREATE_VCPU: | |
2893 | r = kvm_vm_ioctl_create_vcpu(kvm, arg); | |
2894 | if (r < 0) | |
2895 | goto out; | |
2896 | break; | |
6aa8b732 AK |
2897 | case KVM_SET_MEMORY_REGION: { |
2898 | struct kvm_memory_region kvm_mem; | |
2899 | ||
2900 | r = -EFAULT; | |
2f366987 | 2901 | if (copy_from_user(&kvm_mem, argp, sizeof kvm_mem)) |
6aa8b732 | 2902 | goto out; |
2c6f5df9 | 2903 | r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_mem); |
6aa8b732 AK |
2904 | if (r) |
2905 | goto out; | |
2906 | break; | |
2907 | } | |
2908 | case KVM_GET_DIRTY_LOG: { | |
2909 | struct kvm_dirty_log log; | |
2910 | ||
2911 | r = -EFAULT; | |
2f366987 | 2912 | if (copy_from_user(&log, argp, sizeof log)) |
6aa8b732 | 2913 | goto out; |
2c6f5df9 | 2914 | r = kvm_vm_ioctl_get_dirty_log(kvm, &log); |
6aa8b732 AK |
2915 | if (r) |
2916 | goto out; | |
2917 | break; | |
2918 | } | |
e8207547 AK |
2919 | case KVM_SET_MEMORY_ALIAS: { |
2920 | struct kvm_memory_alias alias; | |
2921 | ||
2922 | r = -EFAULT; | |
2923 | if (copy_from_user(&alias, argp, sizeof alias)) | |
2924 | goto out; | |
2925 | r = kvm_vm_ioctl_set_memory_alias(kvm, &alias); | |
2926 | if (r) | |
2927 | goto out; | |
2928 | break; | |
2929 | } | |
85f455f7 ED |
2930 | case KVM_CREATE_IRQCHIP: |
2931 | r = -ENOMEM; | |
2932 | kvm->vpic = kvm_create_pic(kvm); | |
1fd4f2a5 ED |
2933 | if (kvm->vpic) { |
2934 | r = kvm_ioapic_init(kvm); | |
2935 | if (r) { | |
2936 | kfree(kvm->vpic); | |
2937 | kvm->vpic = NULL; | |
2938 | goto out; | |
2939 | } | |
2940 | } | |
85f455f7 ED |
2941 | else |
2942 | goto out; | |
2943 | break; | |
2944 | case KVM_IRQ_LINE: { | |
2945 | struct kvm_irq_level irq_event; | |
2946 | ||
2947 | r = -EFAULT; | |
2948 | if (copy_from_user(&irq_event, argp, sizeof irq_event)) | |
2949 | goto out; | |
2950 | if (irqchip_in_kernel(kvm)) { | |
9cf98828 | 2951 | mutex_lock(&kvm->lock); |
85f455f7 ED |
2952 | if (irq_event.irq < 16) |
2953 | kvm_pic_set_irq(pic_irqchip(kvm), | |
2954 | irq_event.irq, | |
2955 | irq_event.level); | |
1fd4f2a5 ED |
2956 | kvm_ioapic_set_irq(kvm->vioapic, |
2957 | irq_event.irq, | |
2958 | irq_event.level); | |
9cf98828 | 2959 | mutex_unlock(&kvm->lock); |
85f455f7 ED |
2960 | r = 0; |
2961 | } | |
2962 | break; | |
2963 | } | |
6ceb9d79 HQ |
2964 | case KVM_GET_IRQCHIP: { |
2965 | /* 0: PIC master, 1: PIC slave, 2: IOAPIC */ | |
2966 | struct kvm_irqchip chip; | |
2967 | ||
2968 | r = -EFAULT; | |
2969 | if (copy_from_user(&chip, argp, sizeof chip)) | |
2970 | goto out; | |
2971 | r = -ENXIO; | |
2972 | if (!irqchip_in_kernel(kvm)) | |
2973 | goto out; | |
2974 | r = kvm_vm_ioctl_get_irqchip(kvm, &chip); | |
2975 | if (r) | |
2976 | goto out; | |
2977 | r = -EFAULT; | |
2978 | if (copy_to_user(argp, &chip, sizeof chip)) | |
2979 | goto out; | |
2980 | r = 0; | |
2981 | break; | |
2982 | } | |
2983 | case KVM_SET_IRQCHIP: { | |
2984 | /* 0: PIC master, 1: PIC slave, 2: IOAPIC */ | |
2985 | struct kvm_irqchip chip; | |
2986 | ||
2987 | r = -EFAULT; | |
2988 | if (copy_from_user(&chip, argp, sizeof chip)) | |
2989 | goto out; | |
2990 | r = -ENXIO; | |
2991 | if (!irqchip_in_kernel(kvm)) | |
2992 | goto out; | |
2993 | r = kvm_vm_ioctl_set_irqchip(kvm, &chip); | |
2994 | if (r) | |
2995 | goto out; | |
2996 | r = 0; | |
2997 | break; | |
2998 | } | |
f17abe9a AK |
2999 | default: |
3000 | ; | |
3001 | } | |
3002 | out: | |
3003 | return r; | |
3004 | } | |
3005 | ||
3006 | static struct page *kvm_vm_nopage(struct vm_area_struct *vma, | |
3007 | unsigned long address, | |
3008 | int *type) | |
3009 | { | |
3010 | struct kvm *kvm = vma->vm_file->private_data; | |
3011 | unsigned long pgoff; | |
f17abe9a AK |
3012 | struct page *page; |
3013 | ||
f17abe9a | 3014 | pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
954bbbc2 | 3015 | page = gfn_to_page(kvm, pgoff); |
f17abe9a AK |
3016 | if (!page) |
3017 | return NOPAGE_SIGBUS; | |
3018 | get_page(page); | |
cd0d9137 NAQ |
3019 | if (type != NULL) |
3020 | *type = VM_FAULT_MINOR; | |
3021 | ||
f17abe9a AK |
3022 | return page; |
3023 | } | |
3024 | ||
3025 | static struct vm_operations_struct kvm_vm_vm_ops = { | |
3026 | .nopage = kvm_vm_nopage, | |
3027 | }; | |
3028 | ||
3029 | static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma) | |
3030 | { | |
3031 | vma->vm_ops = &kvm_vm_vm_ops; | |
3032 | return 0; | |
3033 | } | |
3034 | ||
3035 | static struct file_operations kvm_vm_fops = { | |
3036 | .release = kvm_vm_release, | |
3037 | .unlocked_ioctl = kvm_vm_ioctl, | |
3038 | .compat_ioctl = kvm_vm_ioctl, | |
3039 | .mmap = kvm_vm_mmap, | |
3040 | }; | |
3041 | ||
3042 | static int kvm_dev_ioctl_create_vm(void) | |
3043 | { | |
3044 | int fd, r; | |
3045 | struct inode *inode; | |
3046 | struct file *file; | |
3047 | struct kvm *kvm; | |
3048 | ||
f17abe9a | 3049 | kvm = kvm_create_vm(); |
d6d28168 AK |
3050 | if (IS_ERR(kvm)) |
3051 | return PTR_ERR(kvm); | |
3052 | r = anon_inode_getfd(&fd, &inode, &file, "kvm-vm", &kvm_vm_fops, kvm); | |
3053 | if (r) { | |
3054 | kvm_destroy_vm(kvm); | |
3055 | return r; | |
f17abe9a AK |
3056 | } |
3057 | ||
bccf2150 | 3058 | kvm->filp = file; |
f17abe9a | 3059 | |
f17abe9a | 3060 | return fd; |
f17abe9a AK |
3061 | } |
3062 | ||
3063 | static long kvm_dev_ioctl(struct file *filp, | |
3064 | unsigned int ioctl, unsigned long arg) | |
3065 | { | |
3066 | void __user *argp = (void __user *)arg; | |
07c45a36 | 3067 | long r = -EINVAL; |
f17abe9a AK |
3068 | |
3069 | switch (ioctl) { | |
3070 | case KVM_GET_API_VERSION: | |
f0fe5108 AK |
3071 | r = -EINVAL; |
3072 | if (arg) | |
3073 | goto out; | |
f17abe9a AK |
3074 | r = KVM_API_VERSION; |
3075 | break; | |
3076 | case KVM_CREATE_VM: | |
f0fe5108 AK |
3077 | r = -EINVAL; |
3078 | if (arg) | |
3079 | goto out; | |
f17abe9a AK |
3080 | r = kvm_dev_ioctl_create_vm(); |
3081 | break; | |
6aa8b732 | 3082 | case KVM_GET_MSR_INDEX_LIST: { |
2f366987 | 3083 | struct kvm_msr_list __user *user_msr_list = argp; |
6aa8b732 AK |
3084 | struct kvm_msr_list msr_list; |
3085 | unsigned n; | |
3086 | ||
3087 | r = -EFAULT; | |
3088 | if (copy_from_user(&msr_list, user_msr_list, sizeof msr_list)) | |
3089 | goto out; | |
3090 | n = msr_list.nmsrs; | |
6f00e68f | 3091 | msr_list.nmsrs = num_msrs_to_save + ARRAY_SIZE(emulated_msrs); |
6aa8b732 AK |
3092 | if (copy_to_user(user_msr_list, &msr_list, sizeof msr_list)) |
3093 | goto out; | |
3094 | r = -E2BIG; | |
bf591b24 | 3095 | if (n < num_msrs_to_save) |
6aa8b732 AK |
3096 | goto out; |
3097 | r = -EFAULT; | |
3098 | if (copy_to_user(user_msr_list->indices, &msrs_to_save, | |
bf591b24 | 3099 | num_msrs_to_save * sizeof(u32))) |
6aa8b732 | 3100 | goto out; |
6f00e68f AK |
3101 | if (copy_to_user(user_msr_list->indices |
3102 | + num_msrs_to_save * sizeof(u32), | |
3103 | &emulated_msrs, | |
3104 | ARRAY_SIZE(emulated_msrs) * sizeof(u32))) | |
3105 | goto out; | |
6aa8b732 | 3106 | r = 0; |
cc1d8955 | 3107 | break; |
6aa8b732 | 3108 | } |
85f455f7 ED |
3109 | case KVM_CHECK_EXTENSION: { |
3110 | int ext = (long)argp; | |
3111 | ||
3112 | switch (ext) { | |
3113 | case KVM_CAP_IRQCHIP: | |
b6958ce4 | 3114 | case KVM_CAP_HLT: |
85f455f7 ED |
3115 | r = 1; |
3116 | break; | |
3117 | default: | |
3118 | r = 0; | |
3119 | break; | |
3120 | } | |
5d308f45 | 3121 | break; |
85f455f7 | 3122 | } |
07c45a36 AK |
3123 | case KVM_GET_VCPU_MMAP_SIZE: |
3124 | r = -EINVAL; | |
3125 | if (arg) | |
3126 | goto out; | |
039576c0 | 3127 | r = 2 * PAGE_SIZE; |
07c45a36 | 3128 | break; |
6aa8b732 AK |
3129 | default: |
3130 | ; | |
3131 | } | |
3132 | out: | |
3133 | return r; | |
3134 | } | |
3135 | ||
6aa8b732 | 3136 | static struct file_operations kvm_chardev_ops = { |
6aa8b732 AK |
3137 | .unlocked_ioctl = kvm_dev_ioctl, |
3138 | .compat_ioctl = kvm_dev_ioctl, | |
6aa8b732 AK |
3139 | }; |
3140 | ||
3141 | static struct miscdevice kvm_dev = { | |
bbe4432e | 3142 | KVM_MINOR, |
6aa8b732 AK |
3143 | "kvm", |
3144 | &kvm_chardev_ops, | |
3145 | }; | |
3146 | ||
774c47f1 AK |
3147 | /* |
3148 | * Make sure that a cpu that is being hot-unplugged does not have any vcpus | |
3149 | * cached on it. | |
3150 | */ | |
3151 | static void decache_vcpus_on_cpu(int cpu) | |
3152 | { | |
3153 | struct kvm *vm; | |
3154 | struct kvm_vcpu *vcpu; | |
3155 | int i; | |
3156 | ||
3157 | spin_lock(&kvm_lock); | |
11ec2804 | 3158 | list_for_each_entry(vm, &vm_list, vm_list) |
774c47f1 | 3159 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { |
fb3f0f51 RR |
3160 | vcpu = vm->vcpus[i]; |
3161 | if (!vcpu) | |
3162 | continue; | |
774c47f1 AK |
3163 | /* |
3164 | * If the vcpu is locked, then it is running on some | |
3165 | * other cpu and therefore it is not cached on the | |
3166 | * cpu in question. | |
3167 | * | |
3168 | * If it's not locked, check the last cpu it executed | |
3169 | * on. | |
3170 | */ | |
3171 | if (mutex_trylock(&vcpu->mutex)) { | |
3172 | if (vcpu->cpu == cpu) { | |
3173 | kvm_arch_ops->vcpu_decache(vcpu); | |
3174 | vcpu->cpu = -1; | |
3175 | } | |
3176 | mutex_unlock(&vcpu->mutex); | |
3177 | } | |
3178 | } | |
3179 | spin_unlock(&kvm_lock); | |
3180 | } | |
3181 | ||
1b6c0168 AK |
3182 | static void hardware_enable(void *junk) |
3183 | { | |
3184 | int cpu = raw_smp_processor_id(); | |
3185 | ||
3186 | if (cpu_isset(cpu, cpus_hardware_enabled)) | |
3187 | return; | |
3188 | cpu_set(cpu, cpus_hardware_enabled); | |
3189 | kvm_arch_ops->hardware_enable(NULL); | |
3190 | } | |
3191 | ||
3192 | static void hardware_disable(void *junk) | |
3193 | { | |
3194 | int cpu = raw_smp_processor_id(); | |
3195 | ||
3196 | if (!cpu_isset(cpu, cpus_hardware_enabled)) | |
3197 | return; | |
3198 | cpu_clear(cpu, cpus_hardware_enabled); | |
3199 | decache_vcpus_on_cpu(cpu); | |
3200 | kvm_arch_ops->hardware_disable(NULL); | |
3201 | } | |
3202 | ||
774c47f1 AK |
3203 | static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val, |
3204 | void *v) | |
3205 | { | |
3206 | int cpu = (long)v; | |
3207 | ||
3208 | switch (val) { | |
cec9ad27 AK |
3209 | case CPU_DYING: |
3210 | case CPU_DYING_FROZEN: | |
6ec8a856 AK |
3211 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
3212 | cpu); | |
3213 | hardware_disable(NULL); | |
3214 | break; | |
774c47f1 | 3215 | case CPU_UP_CANCELED: |
8bb78442 | 3216 | case CPU_UP_CANCELED_FROZEN: |
43934a38 JK |
3217 | printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n", |
3218 | cpu); | |
1b6c0168 | 3219 | smp_call_function_single(cpu, hardware_disable, NULL, 0, 1); |
774c47f1 | 3220 | break; |
43934a38 | 3221 | case CPU_ONLINE: |
8bb78442 | 3222 | case CPU_ONLINE_FROZEN: |
43934a38 JK |
3223 | printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n", |
3224 | cpu); | |
1b6c0168 | 3225 | smp_call_function_single(cpu, hardware_enable, NULL, 0, 1); |
774c47f1 AK |
3226 | break; |
3227 | } | |
3228 | return NOTIFY_OK; | |
3229 | } | |
3230 | ||
9a2b85c6 RR |
3231 | static int kvm_reboot(struct notifier_block *notifier, unsigned long val, |
3232 | void *v) | |
3233 | { | |
3234 | if (val == SYS_RESTART) { | |
3235 | /* | |
3236 | * Some (well, at least mine) BIOSes hang on reboot if | |
3237 | * in vmx root mode. | |
3238 | */ | |
3239 | printk(KERN_INFO "kvm: exiting hardware virtualization\n"); | |
3240 | on_each_cpu(hardware_disable, NULL, 0, 1); | |
3241 | } | |
3242 | return NOTIFY_OK; | |
3243 | } | |
3244 | ||
3245 | static struct notifier_block kvm_reboot_notifier = { | |
3246 | .notifier_call = kvm_reboot, | |
3247 | .priority = 0, | |
3248 | }; | |
3249 | ||
2eeb2e94 GH |
3250 | void kvm_io_bus_init(struct kvm_io_bus *bus) |
3251 | { | |
3252 | memset(bus, 0, sizeof(*bus)); | |
3253 | } | |
3254 | ||
3255 | void kvm_io_bus_destroy(struct kvm_io_bus *bus) | |
3256 | { | |
3257 | int i; | |
3258 | ||
3259 | for (i = 0; i < bus->dev_count; i++) { | |
3260 | struct kvm_io_device *pos = bus->devs[i]; | |
3261 | ||
3262 | kvm_iodevice_destructor(pos); | |
3263 | } | |
3264 | } | |
3265 | ||
3266 | struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr) | |
3267 | { | |
3268 | int i; | |
3269 | ||
3270 | for (i = 0; i < bus->dev_count; i++) { | |
3271 | struct kvm_io_device *pos = bus->devs[i]; | |
3272 | ||
3273 | if (pos->in_range(pos, addr)) | |
3274 | return pos; | |
3275 | } | |
3276 | ||
3277 | return NULL; | |
3278 | } | |
3279 | ||
3280 | void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev) | |
3281 | { | |
3282 | BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1)); | |
3283 | ||
3284 | bus->devs[bus->dev_count++] = dev; | |
3285 | } | |
3286 | ||
774c47f1 AK |
3287 | static struct notifier_block kvm_cpu_notifier = { |
3288 | .notifier_call = kvm_cpu_hotplug, | |
3289 | .priority = 20, /* must be > scheduler priority */ | |
3290 | }; | |
3291 | ||
1165f5fe AK |
3292 | static u64 stat_get(void *_offset) |
3293 | { | |
3294 | unsigned offset = (long)_offset; | |
3295 | u64 total = 0; | |
3296 | struct kvm *kvm; | |
3297 | struct kvm_vcpu *vcpu; | |
3298 | int i; | |
3299 | ||
3300 | spin_lock(&kvm_lock); | |
3301 | list_for_each_entry(kvm, &vm_list, vm_list) | |
3302 | for (i = 0; i < KVM_MAX_VCPUS; ++i) { | |
fb3f0f51 RR |
3303 | vcpu = kvm->vcpus[i]; |
3304 | if (vcpu) | |
3305 | total += *(u32 *)((void *)vcpu + offset); | |
1165f5fe AK |
3306 | } |
3307 | spin_unlock(&kvm_lock); | |
3308 | return total; | |
3309 | } | |
3310 | ||
3dea7ca7 | 3311 | DEFINE_SIMPLE_ATTRIBUTE(stat_fops, stat_get, NULL, "%llu\n"); |
1165f5fe | 3312 | |
6aa8b732 AK |
3313 | static __init void kvm_init_debug(void) |
3314 | { | |
3315 | struct kvm_stats_debugfs_item *p; | |
3316 | ||
8b6d44c7 | 3317 | debugfs_dir = debugfs_create_dir("kvm", NULL); |
6aa8b732 | 3318 | for (p = debugfs_entries; p->name; ++p) |
1165f5fe AK |
3319 | p->dentry = debugfs_create_file(p->name, 0444, debugfs_dir, |
3320 | (void *)(long)p->offset, | |
3321 | &stat_fops); | |
6aa8b732 AK |
3322 | } |
3323 | ||
3324 | static void kvm_exit_debug(void) | |
3325 | { | |
3326 | struct kvm_stats_debugfs_item *p; | |
3327 | ||
3328 | for (p = debugfs_entries; p->name; ++p) | |
3329 | debugfs_remove(p->dentry); | |
3330 | debugfs_remove(debugfs_dir); | |
3331 | } | |
3332 | ||
59ae6c6b AK |
3333 | static int kvm_suspend(struct sys_device *dev, pm_message_t state) |
3334 | { | |
4267c41a | 3335 | hardware_disable(NULL); |
59ae6c6b AK |
3336 | return 0; |
3337 | } | |
3338 | ||
3339 | static int kvm_resume(struct sys_device *dev) | |
3340 | { | |
4267c41a | 3341 | hardware_enable(NULL); |
59ae6c6b AK |
3342 | return 0; |
3343 | } | |
3344 | ||
3345 | static struct sysdev_class kvm_sysdev_class = { | |
3346 | set_kset_name("kvm"), | |
3347 | .suspend = kvm_suspend, | |
3348 | .resume = kvm_resume, | |
3349 | }; | |
3350 | ||
3351 | static struct sys_device kvm_sysdev = { | |
3352 | .id = 0, | |
3353 | .cls = &kvm_sysdev_class, | |
3354 | }; | |
3355 | ||
6aa8b732 AK |
3356 | hpa_t bad_page_address; |
3357 | ||
15ad7146 AK |
3358 | static inline |
3359 | struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn) | |
3360 | { | |
3361 | return container_of(pn, struct kvm_vcpu, preempt_notifier); | |
3362 | } | |
3363 | ||
3364 | static void kvm_sched_in(struct preempt_notifier *pn, int cpu) | |
3365 | { | |
3366 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
3367 | ||
3368 | kvm_arch_ops->vcpu_load(vcpu, cpu); | |
3369 | } | |
3370 | ||
3371 | static void kvm_sched_out(struct preempt_notifier *pn, | |
3372 | struct task_struct *next) | |
3373 | { | |
3374 | struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn); | |
3375 | ||
3376 | kvm_arch_ops->vcpu_put(vcpu); | |
3377 | } | |
3378 | ||
c16f862d RR |
3379 | int kvm_init_arch(struct kvm_arch_ops *ops, unsigned int vcpu_size, |
3380 | struct module *module) | |
6aa8b732 AK |
3381 | { |
3382 | int r; | |
002c7f7c | 3383 | int cpu; |
6aa8b732 | 3384 | |
09db28b8 YI |
3385 | if (kvm_arch_ops) { |
3386 | printk(KERN_ERR "kvm: already loaded the other module\n"); | |
3387 | return -EEXIST; | |
3388 | } | |
3389 | ||
e097f35c | 3390 | if (!ops->cpu_has_kvm_support()) { |
6aa8b732 AK |
3391 | printk(KERN_ERR "kvm: no hardware support\n"); |
3392 | return -EOPNOTSUPP; | |
3393 | } | |
e097f35c | 3394 | if (ops->disabled_by_bios()) { |
6aa8b732 AK |
3395 | printk(KERN_ERR "kvm: disabled by bios\n"); |
3396 | return -EOPNOTSUPP; | |
3397 | } | |
3398 | ||
e097f35c YI |
3399 | kvm_arch_ops = ops; |
3400 | ||
6aa8b732 AK |
3401 | r = kvm_arch_ops->hardware_setup(); |
3402 | if (r < 0) | |
ca45aaae | 3403 | goto out; |
6aa8b732 | 3404 | |
002c7f7c YS |
3405 | for_each_online_cpu(cpu) { |
3406 | smp_call_function_single(cpu, | |
3407 | kvm_arch_ops->check_processor_compatibility, | |
3408 | &r, 0, 1); | |
3409 | if (r < 0) | |
3410 | goto out_free_0; | |
3411 | } | |
3412 | ||
1b6c0168 | 3413 | on_each_cpu(hardware_enable, NULL, 0, 1); |
774c47f1 AK |
3414 | r = register_cpu_notifier(&kvm_cpu_notifier); |
3415 | if (r) | |
3416 | goto out_free_1; | |
6aa8b732 AK |
3417 | register_reboot_notifier(&kvm_reboot_notifier); |
3418 | ||
59ae6c6b AK |
3419 | r = sysdev_class_register(&kvm_sysdev_class); |
3420 | if (r) | |
3421 | goto out_free_2; | |
3422 | ||
3423 | r = sysdev_register(&kvm_sysdev); | |
3424 | if (r) | |
3425 | goto out_free_3; | |
3426 | ||
c16f862d RR |
3427 | /* A kmem cache lets us meet the alignment requirements of fx_save. */ |
3428 | kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, | |
3429 | __alignof__(struct kvm_vcpu), 0, 0); | |
3430 | if (!kvm_vcpu_cache) { | |
3431 | r = -ENOMEM; | |
3432 | goto out_free_4; | |
3433 | } | |
3434 | ||
6aa8b732 AK |
3435 | kvm_chardev_ops.owner = module; |
3436 | ||
3437 | r = misc_register(&kvm_dev); | |
3438 | if (r) { | |
3439 | printk (KERN_ERR "kvm: misc device register failed\n"); | |
3440 | goto out_free; | |
3441 | } | |
3442 | ||
15ad7146 AK |
3443 | kvm_preempt_ops.sched_in = kvm_sched_in; |
3444 | kvm_preempt_ops.sched_out = kvm_sched_out; | |
3445 | ||
6aa8b732 AK |
3446 | return r; |
3447 | ||
3448 | out_free: | |
c16f862d RR |
3449 | kmem_cache_destroy(kvm_vcpu_cache); |
3450 | out_free_4: | |
59ae6c6b AK |
3451 | sysdev_unregister(&kvm_sysdev); |
3452 | out_free_3: | |
3453 | sysdev_class_unregister(&kvm_sysdev_class); | |
3454 | out_free_2: | |
6aa8b732 | 3455 | unregister_reboot_notifier(&kvm_reboot_notifier); |
774c47f1 AK |
3456 | unregister_cpu_notifier(&kvm_cpu_notifier); |
3457 | out_free_1: | |
1b6c0168 | 3458 | on_each_cpu(hardware_disable, NULL, 0, 1); |
002c7f7c | 3459 | out_free_0: |
6aa8b732 | 3460 | kvm_arch_ops->hardware_unsetup(); |
ca45aaae AK |
3461 | out: |
3462 | kvm_arch_ops = NULL; | |
6aa8b732 AK |
3463 | return r; |
3464 | } | |
3465 | ||
3466 | void kvm_exit_arch(void) | |
3467 | { | |
3468 | misc_deregister(&kvm_dev); | |
c16f862d | 3469 | kmem_cache_destroy(kvm_vcpu_cache); |
59ae6c6b AK |
3470 | sysdev_unregister(&kvm_sysdev); |
3471 | sysdev_class_unregister(&kvm_sysdev_class); | |
6aa8b732 | 3472 | unregister_reboot_notifier(&kvm_reboot_notifier); |
59ae6c6b | 3473 | unregister_cpu_notifier(&kvm_cpu_notifier); |
1b6c0168 | 3474 | on_each_cpu(hardware_disable, NULL, 0, 1); |
6aa8b732 | 3475 | kvm_arch_ops->hardware_unsetup(); |
09db28b8 | 3476 | kvm_arch_ops = NULL; |
6aa8b732 AK |
3477 | } |
3478 | ||
3479 | static __init int kvm_init(void) | |
3480 | { | |
3481 | static struct page *bad_page; | |
37e29d90 AK |
3482 | int r; |
3483 | ||
b5a33a75 AK |
3484 | r = kvm_mmu_module_init(); |
3485 | if (r) | |
3486 | goto out4; | |
3487 | ||
6aa8b732 AK |
3488 | kvm_init_debug(); |
3489 | ||
bf591b24 MR |
3490 | kvm_init_msr_list(); |
3491 | ||
6aa8b732 AK |
3492 | if ((bad_page = alloc_page(GFP_KERNEL)) == NULL) { |
3493 | r = -ENOMEM; | |
3494 | goto out; | |
3495 | } | |
3496 | ||
3497 | bad_page_address = page_to_pfn(bad_page) << PAGE_SHIFT; | |
3498 | memset(__va(bad_page_address), 0, PAGE_SIZE); | |
3499 | ||
58e690e6 | 3500 | return 0; |
6aa8b732 AK |
3501 | |
3502 | out: | |
3503 | kvm_exit_debug(); | |
b5a33a75 AK |
3504 | kvm_mmu_module_exit(); |
3505 | out4: | |
6aa8b732 AK |
3506 | return r; |
3507 | } | |
3508 | ||
3509 | static __exit void kvm_exit(void) | |
3510 | { | |
3511 | kvm_exit_debug(); | |
3512 | __free_page(pfn_to_page(bad_page_address >> PAGE_SHIFT)); | |
b5a33a75 | 3513 | kvm_mmu_module_exit(); |
6aa8b732 AK |
3514 | } |
3515 | ||
3516 | module_init(kvm_init) | |
3517 | module_exit(kvm_exit) | |
3518 | ||
3519 | EXPORT_SYMBOL_GPL(kvm_init_arch); | |
3520 | EXPORT_SYMBOL_GPL(kvm_exit_arch); |