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