projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[PATCH] KVM: MMU: Zap shadow page table entries on writes to guest page tables
[deliverable/linux.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"
19
20#include <linux/kvm.h>
21#include <linux/module.h>
22#include <linux/errno.h>
23#include <asm/processor.h>
24#include <linux/percpu.h>
25#include <linux/gfp.h>
26#include <asm/msr.h>
27#include <linux/mm.h>
28#include <linux/miscdevice.h>
29#include <linux/vmalloc.h>
30#include <asm/uaccess.h>
31#include <linux/reboot.h>
32#include <asm/io.h>
33#include <linux/debugfs.h>
34#include <linux/highmem.h>
35#include <linux/file.h>
36#include <asm/desc.h>
37
38#include "x86_emulate.h"
39#include "segment_descriptor.h"
40
41MODULE_AUTHOR("Qumranet");
42MODULE_LICENSE("GPL");
43
44struct kvm_arch_ops *kvm_arch_ops;
45struct kvm_stat kvm_stat;
46EXPORT_SYMBOL_GPL(kvm_stat);
47
48static struct kvm_stats_debugfs_item {
49 const char *name;
50 u32 *data;
51 struct dentry *dentry;
52} debugfs_entries[] = {
53 { "pf_fixed", &kvm_stat.pf_fixed },
54 { "pf_guest", &kvm_stat.pf_guest },
55 { "tlb_flush", &kvm_stat.tlb_flush },
56 { "invlpg", &kvm_stat.invlpg },
57 { "exits", &kvm_stat.exits },
58 { "io_exits", &kvm_stat.io_exits },
59 { "mmio_exits", &kvm_stat.mmio_exits },
60 { "signal_exits", &kvm_stat.signal_exits },
c1150d8c
DL		 61 { "irq_window", &kvm_stat.irq_window_exits },
62 { "halt_exits", &kvm_stat.halt_exits },
63 { "request_irq", &kvm_stat.request_irq_exits },
6aa8b732
AK		 64 { "irq_exits", &kvm_stat.irq_exits },
65 { 0, 0 }
66};
67
68static struct dentry *debugfs_dir;
69
70#define MAX_IO_MSRS 256
71
72#define CR0_RESEVED_BITS 0xffffffff1ffaffc0ULL
73#define LMSW_GUEST_MASK 0x0eULL
74#define CR4_RESEVED_BITS (~((1ULL << 11) - 1))
75#define CR8_RESEVED_BITS (~0x0fULL)
76#define EFER_RESERVED_BITS 0xfffffffffffff2fe
77
05b3e0c2 78#ifdef CONFIG_X86_64
6aa8b732
AK		 79// LDT or TSS descriptor in the GDT. 16 bytes.
80struct segment_descriptor_64 {
81 struct segment_descriptor s;
82 u32 base_higher;
83 u32 pad_zero;
84};
85
86#endif
87
88unsigned long segment_base(u16 selector)
89{
90 struct descriptor_table gdt;
91 struct segment_descriptor *d;
92 unsigned long table_base;
93 typedef unsigned long ul;
94 unsigned long v;
95
96 if (selector == 0)
97 return 0;
98
99 asm ("sgdt %0" : "=m"(gdt));
100 table_base = gdt.base;
101
102 if (selector & 4) { /* from ldt */
103 u16 ldt_selector;
104
105 asm ("sldt %0" : "=g"(ldt_selector));
106 table_base = segment_base(ldt_selector);
107 }
108 d = (struct segment_descriptor *)(table_base + (selector & ~7));
109 v = d->base_low | ((ul)d->base_mid << 16) | ((ul)d->base_high << 24);
05b3e0c2 110#ifdef CONFIG_X86_64
6aa8b732
AK		 111 if (d->system == 0
112 && (d->type == 2 || d->type == 9 || d->type == 11))
113 v |= ((ul)((struct segment_descriptor_64 *)d)->base_higher) << 32;
114#endif
115 return v;
116}
117EXPORT_SYMBOL_GPL(segment_base);
118
5aacf0ca
JM		 119static inline int valid_vcpu(int n)
120{
121 return likely(n >= 0 && n < KVM_MAX_VCPUS);
122}
123
6aa8b732
AK		 124int kvm_read_guest(struct kvm_vcpu *vcpu,
125 gva_t addr,
126 unsigned long size,
127 void *dest)
128{
129 unsigned char *host_buf = dest;
130 unsigned long req_size = size;
131
132 while (size) {
133 hpa_t paddr;
134 unsigned now;
135 unsigned offset;
136 hva_t guest_buf;
137
138 paddr = gva_to_hpa(vcpu, addr);
139
140 if (is_error_hpa(paddr))
141 break;
142
143 guest_buf = (hva_t)kmap_atomic(
144 pfn_to_page(paddr >> PAGE_SHIFT),
145 KM_USER0);
146 offset = addr & ~PAGE_MASK;
147 guest_buf |= offset;
148 now = min(size, PAGE_SIZE - offset);
149 memcpy(host_buf, (void*)guest_buf, now);
150 host_buf += now;
151 addr += now;
152 size -= now;
153 kunmap_atomic((void *)(guest_buf & PAGE_MASK), KM_USER0);
154 }
155 return req_size - size;
156}
157EXPORT_SYMBOL_GPL(kvm_read_guest);
158
159int kvm_write_guest(struct kvm_vcpu *vcpu,
160 gva_t addr,
161 unsigned long size,
162 void *data)
163{
164 unsigned char *host_buf = data;
165 unsigned long req_size = size;
166
167 while (size) {
168 hpa_t paddr;
169 unsigned now;
170 unsigned offset;
171 hva_t guest_buf;
172
173 paddr = gva_to_hpa(vcpu, addr);
174
175 if (is_error_hpa(paddr))
176 break;
177
178 guest_buf = (hva_t)kmap_atomic(
179 pfn_to_page(paddr >> PAGE_SHIFT), KM_USER0);
180 offset = addr & ~PAGE_MASK;
181 guest_buf |= offset;
182 now = min(size, PAGE_SIZE - offset);
183 memcpy((void*)guest_buf, host_buf, now);
184 host_buf += now;
185 addr += now;
186 size -= now;
187 kunmap_atomic((void *)(guest_buf & PAGE_MASK), KM_USER0);
188 }
189 return req_size - size;
190}
191EXPORT_SYMBOL_GPL(kvm_write_guest);
192
193static int vcpu_slot(struct kvm_vcpu *vcpu)
194{
195 return vcpu - vcpu->kvm->vcpus;
196}
197
198/*
199 * Switches to specified vcpu, until a matching vcpu_put()
200 */
201static struct kvm_vcpu *vcpu_load(struct kvm *kvm, int vcpu_slot)
202{
203 struct kvm_vcpu *vcpu = &kvm->vcpus[vcpu_slot];
204
205 mutex_lock(&vcpu->mutex);
206 if (unlikely(!vcpu->vmcs)) {
207 mutex_unlock(&vcpu->mutex);
208 return 0;
209 }
210 return kvm_arch_ops->vcpu_load(vcpu);
211}
212
213static void vcpu_put(struct kvm_vcpu *vcpu)
214{
215 kvm_arch_ops->vcpu_put(vcpu);
6aa8b732
AK		 216 mutex_unlock(&vcpu->mutex);
217}
218
219static int kvm_dev_open(struct inode *inode, struct file *filp)
220{
221 struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
222 int i;
223
224 if (!kvm)
225 return -ENOMEM;
226
227 spin_lock_init(&kvm->lock);
228 INIT_LIST_HEAD(&kvm->active_mmu_pages);
229 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
230 struct kvm_vcpu *vcpu = &kvm->vcpus[i];
231
232 mutex_init(&vcpu->mutex);
233 vcpu->mmu.root_hpa = INVALID_PAGE;
234 INIT_LIST_HEAD(&vcpu->free_pages);
235 }
236 filp->private_data = kvm;
237 return 0;
238}
239
240/*
241 * Free any memory in @free but not in @dont.
242 */
243static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
244 struct kvm_memory_slot *dont)
245{
246 int i;
247
248 if (!dont || free->phys_mem != dont->phys_mem)
249 if (free->phys_mem) {
250 for (i = 0; i < free->npages; ++i)
55a54f79
AK		 251 if (free->phys_mem[i])
252 __free_page(free->phys_mem[i]);
6aa8b732
AK		 253 vfree(free->phys_mem);
254 }
255
256 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
257 vfree(free->dirty_bitmap);
258
259 free->phys_mem = 0;
260 free->npages = 0;
261 free->dirty_bitmap = 0;
262}
263
264static void kvm_free_physmem(struct kvm *kvm)
265{
266 int i;
267
268 for (i = 0; i < kvm->nmemslots; ++i)
269 kvm_free_physmem_slot(&kvm->memslots[i], 0);
270}
271
272static void kvm_free_vcpu(struct kvm_vcpu *vcpu)
273{
274 kvm_arch_ops->vcpu_free(vcpu);
275 kvm_mmu_destroy(vcpu);
276}
277
278static void kvm_free_vcpus(struct kvm *kvm)
279{
280 unsigned int i;
281
282 for (i = 0; i < KVM_MAX_VCPUS; ++i)
283 kvm_free_vcpu(&kvm->vcpus[i]);
284}
285
286static int kvm_dev_release(struct inode *inode, struct file *filp)
287{
288 struct kvm *kvm = filp->private_data;
289
290 kvm_free_vcpus(kvm);
291 kvm_free_physmem(kvm);
292 kfree(kvm);
293 return 0;
294}
295
296static void inject_gp(struct kvm_vcpu *vcpu)
297{
298 kvm_arch_ops->inject_gp(vcpu, 0);
299}
300
1342d353
AK		 301/*
302 * Load the pae pdptrs. Return true is they are all valid.
303 */
304static int load_pdptrs(struct kvm_vcpu *vcpu, unsigned long cr3)
6aa8b732
AK		 305{
306 gfn_t pdpt_gfn = cr3 >> PAGE_SHIFT;
1342d353 307 unsigned offset = ((cr3 & (PAGE_SIZE-1)) >> 5) << 2;
6aa8b732
AK		 308 int i;
309 u64 pdpte;
310 u64 *pdpt;
1342d353 311 int ret;
6aa8b732
AK		 312 struct kvm_memory_slot *memslot;
313
314 spin_lock(&vcpu->kvm->lock);
315 memslot = gfn_to_memslot(vcpu->kvm, pdpt_gfn);
316 /* FIXME: !memslot - emulate? 0xff? */
317 pdpt = kmap_atomic(gfn_to_page(memslot, pdpt_gfn), KM_USER0);
318
1342d353 319 ret = 1;
6aa8b732
AK		 320 for (i = 0; i < 4; ++i) {
321 pdpte = pdpt[offset + i];
1342d353
AK		 322 if ((pdpte & 1) && (pdpte & 0xfffffff0000001e6ull)) {
323 ret = 0;
324 goto out;
325 }
6aa8b732
AK		 326 }
327
1342d353
AK		 328 for (i = 0; i < 4; ++i)
329 vcpu->pdptrs[i] = pdpt[offset + i];
330
331out:
6aa8b732
AK		 332 kunmap_atomic(pdpt, KM_USER0);
333 spin_unlock(&vcpu->kvm->lock);
334
1342d353 335 return ret;
6aa8b732
AK		 336}
337
338void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
339{
340 if (cr0 & CR0_RESEVED_BITS) {
341 printk(KERN_DEBUG "set_cr0: 0x%lx #GP, reserved bits 0x%lx\n",
342 cr0, vcpu->cr0);
343 inject_gp(vcpu);
344 return;
345 }
346
347 if ((cr0 & CR0_NW_MASK) && !(cr0 & CR0_CD_MASK)) {
348 printk(KERN_DEBUG "set_cr0: #GP, CD == 0 && NW == 1\n");
349 inject_gp(vcpu);
350 return;
351 }
352
353 if ((cr0 & CR0_PG_MASK) && !(cr0 & CR0_PE_MASK)) {
354 printk(KERN_DEBUG "set_cr0: #GP, set PG flag "
355 "and a clear PE flag\n");
356 inject_gp(vcpu);
357 return;
358 }
359
360 if (!is_paging(vcpu) && (cr0 & CR0_PG_MASK)) {
05b3e0c2 361#ifdef CONFIG_X86_64
6aa8b732
AK		 362 if ((vcpu->shadow_efer & EFER_LME)) {
363 int cs_db, cs_l;
364
365 if (!is_pae(vcpu)) {
366 printk(KERN_DEBUG "set_cr0: #GP, start paging "
367 "in long mode while PAE is disabled\n");
368 inject_gp(vcpu);
369 return;
370 }
371 kvm_arch_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
372 if (cs_l) {
373 printk(KERN_DEBUG "set_cr0: #GP, start paging "
374 "in long mode while CS.L == 1\n");
375 inject_gp(vcpu);
376 return;
377
378 }
379 } else
380#endif
1342d353 381 if (is_pae(vcpu) && !load_pdptrs(vcpu, vcpu->cr3)) {
6aa8b732
AK		 382 printk(KERN_DEBUG "set_cr0: #GP, pdptrs "
383 "reserved bits\n");
384 inject_gp(vcpu);
385 return;
386 }
387
388 }
389
390 kvm_arch_ops->set_cr0(vcpu, cr0);
391 vcpu->cr0 = cr0;
392
393 spin_lock(&vcpu->kvm->lock);
394 kvm_mmu_reset_context(vcpu);
395 spin_unlock(&vcpu->kvm->lock);
396 return;
397}
398EXPORT_SYMBOL_GPL(set_cr0);
399
400void lmsw(struct kvm_vcpu *vcpu, unsigned long msw)
401{
399badf3 402 kvm_arch_ops->decache_cr0_cr4_guest_bits(vcpu);
6aa8b732
AK		 403 set_cr0(vcpu, (vcpu->cr0 & ~0x0ful) | (msw & 0x0f));
404}
405EXPORT_SYMBOL_GPL(lmsw);
406
407void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
408{
409 if (cr4 & CR4_RESEVED_BITS) {
410 printk(KERN_DEBUG "set_cr4: #GP, reserved bits\n");
411 inject_gp(vcpu);
412 return;
413 }
414
a9058ecd 415 if (is_long_mode(vcpu)) {
6aa8b732
AK		 416 if (!(cr4 & CR4_PAE_MASK)) {
417 printk(KERN_DEBUG "set_cr4: #GP, clearing PAE while "
418 "in long mode\n");
419 inject_gp(vcpu);
420 return;
421 }
422 } else if (is_paging(vcpu) && !is_pae(vcpu) && (cr4 & CR4_PAE_MASK)
1342d353 423 && !load_pdptrs(vcpu, vcpu->cr3)) {
6aa8b732
AK		 424 printk(KERN_DEBUG "set_cr4: #GP, pdptrs reserved bits\n");
425 inject_gp(vcpu);
426 }
427
428 if (cr4 & CR4_VMXE_MASK) {
429 printk(KERN_DEBUG "set_cr4: #GP, setting VMXE\n");
430 inject_gp(vcpu);
431 return;
432 }
433 kvm_arch_ops->set_cr4(vcpu, cr4);
434 spin_lock(&vcpu->kvm->lock);
435 kvm_mmu_reset_context(vcpu);
436 spin_unlock(&vcpu->kvm->lock);
437}
438EXPORT_SYMBOL_GPL(set_cr4);
439
440void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
441{
a9058ecd 442 if (is_long_mode(vcpu)) {
6aa8b732
AK		 443 if ( cr3 & CR3_L_MODE_RESEVED_BITS) {
444 printk(KERN_DEBUG "set_cr3: #GP, reserved bits\n");
445 inject_gp(vcpu);
446 return;
447 }
448 } else {
449 if (cr3 & CR3_RESEVED_BITS) {
450 printk(KERN_DEBUG "set_cr3: #GP, reserved bits\n");
451 inject_gp(vcpu);
452 return;
453 }
454 if (is_paging(vcpu) && is_pae(vcpu) &&
1342d353 455 !load_pdptrs(vcpu, cr3)) {
6aa8b732
AK		 456 printk(KERN_DEBUG "set_cr3: #GP, pdptrs "
457 "reserved bits\n");
458 inject_gp(vcpu);
459 return;
460 }
461 }
462
463 vcpu->cr3 = cr3;
464 spin_lock(&vcpu->kvm->lock);
465 vcpu->mmu.new_cr3(vcpu);
466 spin_unlock(&vcpu->kvm->lock);
467}
468EXPORT_SYMBOL_GPL(set_cr3);
469
470void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8)
471{
472 if ( cr8 & CR8_RESEVED_BITS) {
473 printk(KERN_DEBUG "set_cr8: #GP, reserved bits 0x%lx\n", cr8);
474 inject_gp(vcpu);
475 return;
476 }
477 vcpu->cr8 = cr8;
478}
479EXPORT_SYMBOL_GPL(set_cr8);
480
481void fx_init(struct kvm_vcpu *vcpu)
482{
483 struct __attribute__ ((__packed__)) fx_image_s {
484 u16 control; //fcw
485 u16 status; //fsw
486 u16 tag; // ftw
487 u16 opcode; //fop
488 u64 ip; // fpu ip
489 u64 operand;// fpu dp
490 u32 mxcsr;
491 u32 mxcsr_mask;
492
493 } *fx_image;
494
495 fx_save(vcpu->host_fx_image);
496 fpu_init();
497 fx_save(vcpu->guest_fx_image);
498 fx_restore(vcpu->host_fx_image);
499
500 fx_image = (struct fx_image_s *)vcpu->guest_fx_image;
501 fx_image->mxcsr = 0x1f80;
502 memset(vcpu->guest_fx_image + sizeof(struct fx_image_s),
503 0, FX_IMAGE_SIZE - sizeof(struct fx_image_s));
504}
505EXPORT_SYMBOL_GPL(fx_init);
506
507/*
508 * Creates some virtual cpus. Good luck creating more than one.
509 */
510static int kvm_dev_ioctl_create_vcpu(struct kvm *kvm, int n)
511{
512 int r;
513 struct kvm_vcpu *vcpu;
514
515 r = -EINVAL;
5aacf0ca 516 if (!valid_vcpu(n))
6aa8b732
AK		 517 goto out;
518
519 vcpu = &kvm->vcpus[n];
520
521 mutex_lock(&vcpu->mutex);
522
523 if (vcpu->vmcs) {
524 mutex_unlock(&vcpu->mutex);
525 return -EEXIST;
526 }
527
528 vcpu->host_fx_image = (char*)ALIGN((hva_t)vcpu->fx_buf,
529 FX_IMAGE_ALIGN);
530 vcpu->guest_fx_image = vcpu->host_fx_image + FX_IMAGE_SIZE;
531
532 vcpu->cpu = -1; /* First load will set up TR */
533 vcpu->kvm = kvm;
534 r = kvm_arch_ops->vcpu_create(vcpu);
535 if (r < 0)
536 goto out_free_vcpus;
537
8018c27b
IM		 538 r = kvm_mmu_create(vcpu);
539 if (r < 0)
540 goto out_free_vcpus;
6aa8b732 541
8018c27b
IM		 542 kvm_arch_ops->vcpu_load(vcpu);
543 r = kvm_mmu_setup(vcpu);
6aa8b732 544 if (r >= 0)
8018c27b 545 r = kvm_arch_ops->vcpu_setup(vcpu);
6aa8b732
AK		 546 vcpu_put(vcpu);
547
548 if (r < 0)
549 goto out_free_vcpus;
550
551 return 0;
552
553out_free_vcpus:
554 kvm_free_vcpu(vcpu);
555 mutex_unlock(&vcpu->mutex);
556out:
557 return r;
558}
559
560/*
561 * Allocate some memory and give it an address in the guest physical address
562 * space.
563 *
564 * Discontiguous memory is allowed, mostly for framebuffers.
565 */
566static int kvm_dev_ioctl_set_memory_region(struct kvm *kvm,
567 struct kvm_memory_region *mem)
568{
569 int r;
570 gfn_t base_gfn;
571 unsigned long npages;
572 unsigned long i;
573 struct kvm_memory_slot *memslot;
574 struct kvm_memory_slot old, new;
575 int memory_config_version;
576
577 r = -EINVAL;
578 /* General sanity checks */
579 if (mem->memory_size & (PAGE_SIZE - 1))
580 goto out;
581 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
582 goto out;
583 if (mem->slot >= KVM_MEMORY_SLOTS)
584 goto out;
585 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
586 goto out;
587
588 memslot = &kvm->memslots[mem->slot];
589 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
590 npages = mem->memory_size >> PAGE_SHIFT;
591
592 if (!npages)
593 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
594
595raced:
596 spin_lock(&kvm->lock);
597
598 memory_config_version = kvm->memory_config_version;
599 new = old = *memslot;
600
601 new.base_gfn = base_gfn;
602 new.npages = npages;
603 new.flags = mem->flags;
604
605 /* Disallow changing a memory slot's size. */
606 r = -EINVAL;
607 if (npages && old.npages && npages != old.npages)
608 goto out_unlock;
609
610 /* Check for overlaps */
611 r = -EEXIST;
612 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
613 struct kvm_memory_slot *s = &kvm->memslots[i];
614
615 if (s == memslot)
616 continue;
617 if (!((base_gfn + npages <= s->base_gfn) ||
618 (base_gfn >= s->base_gfn + s->npages)))
619 goto out_unlock;
620 }
621 /*
622 * Do memory allocations outside lock. memory_config_version will
623 * detect any races.
624 */
625 spin_unlock(&kvm->lock);
626
627 /* Deallocate if slot is being removed */
628 if (!npages)
629 new.phys_mem = 0;
630
631 /* Free page dirty bitmap if unneeded */
632 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
633 new.dirty_bitmap = 0;
634
635 r = -ENOMEM;
636
637 /* Allocate if a slot is being created */
638 if (npages && !new.phys_mem) {
639 new.phys_mem = vmalloc(npages * sizeof(struct page *));
640
641 if (!new.phys_mem)
642 goto out_free;
643
644 memset(new.phys_mem, 0, npages * sizeof(struct page *));
645 for (i = 0; i < npages; ++i) {
646 new.phys_mem[i] = alloc_page(GFP_HIGHUSER
647 | __GFP_ZERO);
648 if (!new.phys_mem[i])
649 goto out_free;
cd4a4e53 650 new.phys_mem[i]->private = 0;
6aa8b732
AK		 651 }
652 }
653
654 /* Allocate page dirty bitmap if needed */
655 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
656 unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8;
657
658 new.dirty_bitmap = vmalloc(dirty_bytes);
659 if (!new.dirty_bitmap)
660 goto out_free;
661 memset(new.dirty_bitmap, 0, dirty_bytes);
662 }
663
664 spin_lock(&kvm->lock);
665
666 if (memory_config_version != kvm->memory_config_version) {
667 spin_unlock(&kvm->lock);
668 kvm_free_physmem_slot(&new, &old);
669 goto raced;
670 }
671
672 r = -EAGAIN;
673 if (kvm->busy)
674 goto out_unlock;
675
676 if (mem->slot >= kvm->nmemslots)
677 kvm->nmemslots = mem->slot + 1;
678
679 *memslot = new;
680 ++kvm->memory_config_version;
681
682 spin_unlock(&kvm->lock);
683
684 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
685 struct kvm_vcpu *vcpu;
686
687 vcpu = vcpu_load(kvm, i);
688 if (!vcpu)
689 continue;
690 kvm_mmu_reset_context(vcpu);
691 vcpu_put(vcpu);
692 }
693
694 kvm_free_physmem_slot(&old, &new);
695 return 0;
696
697out_unlock:
698 spin_unlock(&kvm->lock);
699out_free:
700 kvm_free_physmem_slot(&new, &old);
701out:
702 return r;
703}
704
705/*
706 * Get (and clear) the dirty memory log for a memory slot.
707 */
708static int kvm_dev_ioctl_get_dirty_log(struct kvm *kvm,
709 struct kvm_dirty_log *log)
710{
711 struct kvm_memory_slot *memslot;
712 int r, i;
713 int n;
714 unsigned long any = 0;
715
716 spin_lock(&kvm->lock);
717
718 /*
719 * Prevent changes to guest memory configuration even while the lock
720 * is not taken.
721 */
722 ++kvm->busy;
723 spin_unlock(&kvm->lock);
724 r = -EINVAL;
725 if (log->slot >= KVM_MEMORY_SLOTS)
726 goto out;
727
728 memslot = &kvm->memslots[log->slot];
729 r = -ENOENT;
730 if (!memslot->dirty_bitmap)
731 goto out;
732
733 n = ALIGN(memslot->npages, 8) / 8;
734
735 for (i = 0; !any && i < n; ++i)
736 any = memslot->dirty_bitmap[i];
737
738 r = -EFAULT;
739 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
740 goto out;
741
742
743 if (any) {
744 spin_lock(&kvm->lock);
745 kvm_mmu_slot_remove_write_access(kvm, log->slot);
746 spin_unlock(&kvm->lock);
747 memset(memslot->dirty_bitmap, 0, n);
748 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
749 struct kvm_vcpu *vcpu = vcpu_load(kvm, i);
750
751 if (!vcpu)
752 continue;
753 kvm_arch_ops->tlb_flush(vcpu);
754 vcpu_put(vcpu);
755 }
756 }
757
758 r = 0;
759
760out:
761 spin_lock(&kvm->lock);
762 --kvm->busy;
763 spin_unlock(&kvm->lock);
764 return r;
765}
766
767struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
768{
769 int i;
770
771 for (i = 0; i < kvm->nmemslots; ++i) {
772 struct kvm_memory_slot *memslot = &kvm->memslots[i];
773
774 if (gfn >= memslot->base_gfn
775 && gfn < memslot->base_gfn + memslot->npages)
776 return memslot;
777 }
778 return 0;
779}
780EXPORT_SYMBOL_GPL(gfn_to_memslot);
781
782void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
783{
784 int i;
785 struct kvm_memory_slot *memslot = 0;
786 unsigned long rel_gfn;
787
788 for (i = 0; i < kvm->nmemslots; ++i) {
789 memslot = &kvm->memslots[i];
790
791 if (gfn >= memslot->base_gfn
792 && gfn < memslot->base_gfn + memslot->npages) {
793
794 if (!memslot || !memslot->dirty_bitmap)
795 return;
796
797 rel_gfn = gfn - memslot->base_gfn;
798
799 /* avoid RMW */
800 if (!test_bit(rel_gfn, memslot->dirty_bitmap))
801 set_bit(rel_gfn, memslot->dirty_bitmap);
802 return;
803 }
804 }
805}
806
807static int emulator_read_std(unsigned long addr,
808 unsigned long *val,
809 unsigned int bytes,
810 struct x86_emulate_ctxt *ctxt)
811{
812 struct kvm_vcpu *vcpu = ctxt->vcpu;
813 void *data = val;
814
815 while (bytes) {
816 gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr);
817 unsigned offset = addr & (PAGE_SIZE-1);
818 unsigned tocopy = min(bytes, (unsigned)PAGE_SIZE - offset);
819 unsigned long pfn;
820 struct kvm_memory_slot *memslot;
821 void *page;
822
823 if (gpa == UNMAPPED_GVA)
824 return X86EMUL_PROPAGATE_FAULT;
825 pfn = gpa >> PAGE_SHIFT;
826 memslot = gfn_to_memslot(vcpu->kvm, pfn);
827 if (!memslot)
828 return X86EMUL_UNHANDLEABLE;
829 page = kmap_atomic(gfn_to_page(memslot, pfn), KM_USER0);
830
831 memcpy(data, page + offset, tocopy);
832
833 kunmap_atomic(page, KM_USER0);
834
835 bytes -= tocopy;
836 data += tocopy;
837 addr += tocopy;
838 }
839
840 return X86EMUL_CONTINUE;
841}
842
843static int emulator_write_std(unsigned long addr,
844 unsigned long val,
845 unsigned int bytes,
846 struct x86_emulate_ctxt *ctxt)
847{
848 printk(KERN_ERR "emulator_write_std: addr %lx n %d\n",
849 addr, bytes);
850 return X86EMUL_UNHANDLEABLE;
851}
852
853static int emulator_read_emulated(unsigned long addr,
854 unsigned long *val,
855 unsigned int bytes,
856 struct x86_emulate_ctxt *ctxt)
857{
858 struct kvm_vcpu *vcpu = ctxt->vcpu;
859
860 if (vcpu->mmio_read_completed) {
861 memcpy(val, vcpu->mmio_data, bytes);
862 vcpu->mmio_read_completed = 0;
863 return X86EMUL_CONTINUE;
864 } else if (emulator_read_std(addr, val, bytes, ctxt)
865 == X86EMUL_CONTINUE)
866 return X86EMUL_CONTINUE;
867 else {
868 gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr);
869 if (gpa == UNMAPPED_GVA)
870 return vcpu_printf(vcpu, "not present\n"), X86EMUL_PROPAGATE_FAULT;
871 vcpu->mmio_needed = 1;
872 vcpu->mmio_phys_addr = gpa;
873 vcpu->mmio_size = bytes;
874 vcpu->mmio_is_write = 0;
875
876 return X86EMUL_UNHANDLEABLE;
877 }
878}
879
da4a00f0
AK		 880static int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
881 unsigned long val, int bytes)
882{
883 struct kvm_memory_slot *m;
884 struct page *page;
885 void *virt;
886
887 if (((gpa + bytes - 1) >> PAGE_SHIFT) != (gpa >> PAGE_SHIFT))
888 return 0;
889 m = gfn_to_memslot(vcpu->kvm, gpa >> PAGE_SHIFT);
890 if (!m)
891 return 0;
892 page = gfn_to_page(m, gpa >> PAGE_SHIFT);
893 kvm_mmu_pre_write(vcpu, gpa, bytes);
894 virt = kmap_atomic(page, KM_USER0);
895 memcpy(virt + offset_in_page(gpa), &val, bytes);
896 kunmap_atomic(virt, KM_USER0);
897 kvm_mmu_post_write(vcpu, gpa, bytes);
898 return 1;
899}
900
6aa8b732
AK		 901static int emulator_write_emulated(unsigned long addr,
902 unsigned long val,
903 unsigned int bytes,
904 struct x86_emulate_ctxt *ctxt)
905{
906 struct kvm_vcpu *vcpu = ctxt->vcpu;
907 gpa_t gpa = vcpu->mmu.gva_to_gpa(vcpu, addr);
908
909 if (gpa == UNMAPPED_GVA)
910 return X86EMUL_PROPAGATE_FAULT;
911
da4a00f0
AK		 912 if (emulator_write_phys(vcpu, gpa, val, bytes))
913 return X86EMUL_CONTINUE;
914
6aa8b732
AK		 915 vcpu->mmio_needed = 1;
916 vcpu->mmio_phys_addr = gpa;
917 vcpu->mmio_size = bytes;
918 vcpu->mmio_is_write = 1;
919 memcpy(vcpu->mmio_data, &val, bytes);
920
921 return X86EMUL_CONTINUE;
922}
923
924static int emulator_cmpxchg_emulated(unsigned long addr,
925 unsigned long old,
926 unsigned long new,
927 unsigned int bytes,
928 struct x86_emulate_ctxt *ctxt)
929{
930 static int reported;
931
932 if (!reported) {
933 reported = 1;
934 printk(KERN_WARNING "kvm: emulating exchange as write\n");
935 }
936 return emulator_write_emulated(addr, new, bytes, ctxt);
937}
938
939static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg)
940{
941 return kvm_arch_ops->get_segment_base(vcpu, seg);
942}
943
944int emulate_invlpg(struct kvm_vcpu *vcpu, gva_t address)
945{
946 spin_lock(&vcpu->kvm->lock);
947 vcpu->mmu.inval_page(vcpu, address);
948 spin_unlock(&vcpu->kvm->lock);
949 kvm_arch_ops->invlpg(vcpu, address);
950 return X86EMUL_CONTINUE;
951}
952
953int emulate_clts(struct kvm_vcpu *vcpu)
954{
399badf3 955 unsigned long cr0;
6aa8b732 956
399badf3
AK		 957 kvm_arch_ops->decache_cr0_cr4_guest_bits(vcpu);
958 cr0 = vcpu->cr0 & ~CR0_TS_MASK;
6aa8b732
AK		 959 kvm_arch_ops->set_cr0(vcpu, cr0);
960 return X86EMUL_CONTINUE;
961}
962
963int emulator_get_dr(struct x86_emulate_ctxt* ctxt, int dr, unsigned long *dest)
964{
965 struct kvm_vcpu *vcpu = ctxt->vcpu;
966
967 switch (dr) {
968 case 0 ... 3:
969 *dest = kvm_arch_ops->get_dr(vcpu, dr);
970 return X86EMUL_CONTINUE;
971 default:
972 printk(KERN_DEBUG "%s: unexpected dr %u\n",
973 __FUNCTION__, dr);
974 return X86EMUL_UNHANDLEABLE;
975 }
976}
977
978int emulator_set_dr(struct x86_emulate_ctxt *ctxt, int dr, unsigned long value)
979{
980 unsigned long mask = (ctxt->mode == X86EMUL_MODE_PROT64) ? ~0ULL : ~0U;
981 int exception;
982
983 kvm_arch_ops->set_dr(ctxt->vcpu, dr, value & mask, &exception);
984 if (exception) {
985 /* FIXME: better handling */
986 return X86EMUL_UNHANDLEABLE;
987 }
988 return X86EMUL_CONTINUE;
989}
990
991static void report_emulation_failure(struct x86_emulate_ctxt *ctxt)
992{
993 static int reported;
994 u8 opcodes[4];
995 unsigned long rip = ctxt->vcpu->rip;
996 unsigned long rip_linear;
997
998 rip_linear = rip + get_segment_base(ctxt->vcpu, VCPU_SREG_CS);
999
1000 if (reported)
1001 return;
1002
1003 emulator_read_std(rip_linear, (void *)opcodes, 4, ctxt);
1004
1005 printk(KERN_ERR "emulation failed but !mmio_needed?"
1006 " rip %lx %02x %02x %02x %02x\n",
1007 rip, opcodes[0], opcodes[1], opcodes[2], opcodes[3]);
1008 reported = 1;
1009}
1010
1011struct x86_emulate_ops emulate_ops = {
1012 .read_std = emulator_read_std,
1013 .write_std = emulator_write_std,
1014 .read_emulated = emulator_read_emulated,
1015 .write_emulated = emulator_write_emulated,
1016 .cmpxchg_emulated = emulator_cmpxchg_emulated,
1017};
1018
1019int emulate_instruction(struct kvm_vcpu *vcpu,
1020 struct kvm_run *run,
1021 unsigned long cr2,
1022 u16 error_code)
1023{
1024 struct x86_emulate_ctxt emulate_ctxt;
1025 int r;
1026 int cs_db, cs_l;
1027
1028 kvm_arch_ops->cache_regs(vcpu);
1029
1030 kvm_arch_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
1031
1032 emulate_ctxt.vcpu = vcpu;
1033 emulate_ctxt.eflags = kvm_arch_ops->get_rflags(vcpu);
1034 emulate_ctxt.cr2 = cr2;
1035 emulate_ctxt.mode = (emulate_ctxt.eflags & X86_EFLAGS_VM)
1036 ? X86EMUL_MODE_REAL : cs_l
1037 ? X86EMUL_MODE_PROT64 : cs_db
1038 ? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16;
1039
1040 if (emulate_ctxt.mode == X86EMUL_MODE_PROT64) {
1041 emulate_ctxt.cs_base = 0;
1042 emulate_ctxt.ds_base = 0;
1043 emulate_ctxt.es_base = 0;
1044 emulate_ctxt.ss_base = 0;
1045 } else {
1046 emulate_ctxt.cs_base = get_segment_base(vcpu, VCPU_SREG_CS);
1047 emulate_ctxt.ds_base = get_segment_base(vcpu, VCPU_SREG_DS);
1048 emulate_ctxt.es_base = get_segment_base(vcpu, VCPU_SREG_ES);
1049 emulate_ctxt.ss_base = get_segment_base(vcpu, VCPU_SREG_SS);
1050 }
1051
1052 emulate_ctxt.gs_base = get_segment_base(vcpu, VCPU_SREG_GS);
1053 emulate_ctxt.fs_base = get_segment_base(vcpu, VCPU_SREG_FS);
1054
1055 vcpu->mmio_is_write = 0;
1056 r = x86_emulate_memop(&emulate_ctxt, &emulate_ops);
1057
1058 if ((r || vcpu->mmio_is_write) && run) {
1059 run->mmio.phys_addr = vcpu->mmio_phys_addr;
1060 memcpy(run->mmio.data, vcpu->mmio_data, 8);
1061 run->mmio.len = vcpu->mmio_size;
1062 run->mmio.is_write = vcpu->mmio_is_write;
1063 }
1064
1065 if (r) {
1066 if (!vcpu->mmio_needed) {
1067 report_emulation_failure(&emulate_ctxt);
1068 return EMULATE_FAIL;
1069 }
1070 return EMULATE_DO_MMIO;
1071 }
1072
1073 kvm_arch_ops->decache_regs(vcpu);
1074 kvm_arch_ops->set_rflags(vcpu, emulate_ctxt.eflags);
1075
1076 if (vcpu->mmio_is_write)
1077 return EMULATE_DO_MMIO;
1078
1079 return EMULATE_DONE;
1080}
1081EXPORT_SYMBOL_GPL(emulate_instruction);
1082
1083static u64 mk_cr_64(u64 curr_cr, u32 new_val)
1084{
1085 return (curr_cr & ~((1ULL << 32) - 1)) | new_val;
1086}
1087
1088void realmode_lgdt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base)
1089{
1090 struct descriptor_table dt = { limit, base };
1091
1092 kvm_arch_ops->set_gdt(vcpu, &dt);
1093}
1094
1095void realmode_lidt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base)
1096{
1097 struct descriptor_table dt = { limit, base };
1098
1099 kvm_arch_ops->set_idt(vcpu, &dt);
1100}
1101
1102void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
1103 unsigned long *rflags)
1104{
1105 lmsw(vcpu, msw);
1106 *rflags = kvm_arch_ops->get_rflags(vcpu);
1107}
1108
1109unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr)
1110{
399badf3 1111 kvm_arch_ops->decache_cr0_cr4_guest_bits(vcpu);
6aa8b732
AK		 1112 switch (cr) {
1113 case 0:
1114 return vcpu->cr0;
1115 case 2:
1116 return vcpu->cr2;
1117 case 3:
1118 return vcpu->cr3;
1119 case 4:
1120 return vcpu->cr4;
1121 default:
1122 vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr);
1123 return 0;
1124 }
1125}
1126
1127void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long val,
1128 unsigned long *rflags)
1129{
1130 switch (cr) {
1131 case 0:
1132 set_cr0(vcpu, mk_cr_64(vcpu->cr0, val));
1133 *rflags = kvm_arch_ops->get_rflags(vcpu);
1134 break;
1135 case 2:
1136 vcpu->cr2 = val;
1137 break;
1138 case 3:
1139 set_cr3(vcpu, val);
1140 break;
1141 case 4:
1142 set_cr4(vcpu, mk_cr_64(vcpu->cr4, val));
1143 break;
1144 default:
1145 vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr);
1146 }
1147}
1148
3bab1f5d
AK		 1149int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata)
1150{
1151 u64 data;
1152
1153 switch (msr) {
1154 case 0xc0010010: /* SYSCFG */
1155 case 0xc0010015: /* HWCR */
1156 case MSR_IA32_PLATFORM_ID:
1157 case MSR_IA32_P5_MC_ADDR:
1158 case MSR_IA32_P5_MC_TYPE:
1159 case MSR_IA32_MC0_CTL:
1160 case MSR_IA32_MCG_STATUS:
1161 case MSR_IA32_MCG_CAP:
1162 case MSR_IA32_MC0_MISC:
1163 case MSR_IA32_MC0_MISC+4:
1164 case MSR_IA32_MC0_MISC+8:
1165 case MSR_IA32_MC0_MISC+12:
1166 case MSR_IA32_MC0_MISC+16:
1167 case MSR_IA32_UCODE_REV:
a8d13ea2 1168 case MSR_IA32_PERF_STATUS:
3bab1f5d
AK		 1169 /* MTRR registers */
1170 case 0xfe:
1171 case 0x200 ... 0x2ff:
1172 data = 0;
1173 break;
a8d13ea2
AK		 1174 case 0xcd: /* fsb frequency */
1175 data = 3;
1176 break;
3bab1f5d
AK		 1177 case MSR_IA32_APICBASE:
1178 data = vcpu->apic_base;
1179 break;
1180#ifdef CONFIG_X86_64
1181 case MSR_EFER:
1182 data = vcpu->shadow_efer;
1183 break;
1184#endif
1185 default:
1186 printk(KERN_ERR "kvm: unhandled rdmsr: 0x%x\n", msr);
1187 return 1;
1188 }
1189 *pdata = data;
1190 return 0;
1191}
1192EXPORT_SYMBOL_GPL(kvm_get_msr_common);
1193
6aa8b732
AK		 1194/*
1195 * Reads an msr value (of 'msr_index') into 'pdata'.
1196 * Returns 0 on success, non-0 otherwise.
1197 * Assumes vcpu_load() was already called.
1198 */
1199static int get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
1200{
1201 return kvm_arch_ops->get_msr(vcpu, msr_index, pdata);
1202}
1203
05b3e0c2 1204#ifdef CONFIG_X86_64
6aa8b732 1205
3bab1f5d 1206static void set_efer(struct kvm_vcpu *vcpu, u64 efer)
6aa8b732 1207{
6aa8b732
AK		 1208 if (efer & EFER_RESERVED_BITS) {
1209 printk(KERN_DEBUG "set_efer: 0x%llx #GP, reserved bits\n",
1210 efer);
1211 inject_gp(vcpu);
1212 return;
1213 }
1214
1215 if (is_paging(vcpu)
1216 && (vcpu->shadow_efer & EFER_LME) != (efer & EFER_LME)) {
1217 printk(KERN_DEBUG "set_efer: #GP, change LME while paging\n");
1218 inject_gp(vcpu);
1219 return;
1220 }
1221
7725f0ba
AK		 1222 kvm_arch_ops->set_efer(vcpu, efer);
1223
6aa8b732
AK		 1224 efer &= ~EFER_LMA;
1225 efer |= vcpu->shadow_efer & EFER_LMA;
1226
1227 vcpu->shadow_efer = efer;
6aa8b732 1228}
6aa8b732
AK		 1229
1230#endif
1231
3bab1f5d
AK		 1232int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data)
1233{
1234 switch (msr) {
1235#ifdef CONFIG_X86_64
1236 case MSR_EFER:
1237 set_efer(vcpu, data);
1238 break;
1239#endif
1240 case MSR_IA32_MC0_STATUS:
1241 printk(KERN_WARNING "%s: MSR_IA32_MC0_STATUS 0x%llx, nop\n",
1242 __FUNCTION__, data);
1243 break;
1244 case MSR_IA32_UCODE_REV:
1245 case MSR_IA32_UCODE_WRITE:
1246 case 0x200 ... 0x2ff: /* MTRRs */
1247 break;
1248 case MSR_IA32_APICBASE:
1249 vcpu->apic_base = data;
1250 break;
1251 default:
1252 printk(KERN_ERR "kvm: unhandled wrmsr: 0x%x\n", msr);
1253 return 1;
1254 }
1255 return 0;
1256}
1257EXPORT_SYMBOL_GPL(kvm_set_msr_common);
1258
6aa8b732
AK		 1259/*
1260 * Writes msr value into into the appropriate "register".
1261 * Returns 0 on success, non-0 otherwise.
1262 * Assumes vcpu_load() was already called.
1263 */
1264static int set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data)
1265{
1266 return kvm_arch_ops->set_msr(vcpu, msr_index, data);
1267}
1268
1269void kvm_resched(struct kvm_vcpu *vcpu)
1270{
1271 vcpu_put(vcpu);
1272 cond_resched();
1273 /* Cannot fail - no vcpu unplug yet. */
1274 vcpu_load(vcpu->kvm, vcpu_slot(vcpu));
1275}
1276EXPORT_SYMBOL_GPL(kvm_resched);
1277
1278void load_msrs(struct vmx_msr_entry *e, int n)
1279{
1280 int i;
1281
1282 for (i = 0; i < n; ++i)
1283 wrmsrl(e[i].index, e[i].data);
1284}
1285EXPORT_SYMBOL_GPL(load_msrs);
1286
1287void save_msrs(struct vmx_msr_entry *e, int n)
1288{
1289 int i;
1290
1291 for (i = 0; i < n; ++i)
1292 rdmsrl(e[i].index, e[i].data);
1293}
1294EXPORT_SYMBOL_GPL(save_msrs);
1295
1296static int kvm_dev_ioctl_run(struct kvm *kvm, struct kvm_run *kvm_run)
1297{
1298 struct kvm_vcpu *vcpu;
1299 int r;
1300
5aacf0ca 1301 if (!valid_vcpu(kvm_run->vcpu))
6aa8b732
AK		 1302 return -EINVAL;
1303
1304 vcpu = vcpu_load(kvm, kvm_run->vcpu);
1305 if (!vcpu)
1306 return -ENOENT;
1307
1308 if (kvm_run->emulated) {
1309 kvm_arch_ops->skip_emulated_instruction(vcpu);
1310 kvm_run->emulated = 0;
1311 }
1312
1313 if (kvm_run->mmio_completed) {
1314 memcpy(vcpu->mmio_data, kvm_run->mmio.data, 8);
1315 vcpu->mmio_read_completed = 1;
1316 }
1317
1318 vcpu->mmio_needed = 0;
1319
1320 r = kvm_arch_ops->run(vcpu, kvm_run);
1321
1322 vcpu_put(vcpu);
1323 return r;
1324}
1325
1326static int kvm_dev_ioctl_get_regs(struct kvm *kvm, struct kvm_regs *regs)
1327{
1328 struct kvm_vcpu *vcpu;
1329
5aacf0ca 1330 if (!valid_vcpu(regs->vcpu))
6aa8b732
AK		 1331 return -EINVAL;
1332
1333 vcpu = vcpu_load(kvm, regs->vcpu);
1334 if (!vcpu)
1335 return -ENOENT;
1336
1337 kvm_arch_ops->cache_regs(vcpu);
1338
1339 regs->rax = vcpu->regs[VCPU_REGS_RAX];
1340 regs->rbx = vcpu->regs[VCPU_REGS_RBX];
1341 regs->rcx = vcpu->regs[VCPU_REGS_RCX];
1342 regs->rdx = vcpu->regs[VCPU_REGS_RDX];
1343 regs->rsi = vcpu->regs[VCPU_REGS_RSI];
1344 regs->rdi = vcpu->regs[VCPU_REGS_RDI];
1345 regs->rsp = vcpu->regs[VCPU_REGS_RSP];
1346 regs->rbp = vcpu->regs[VCPU_REGS_RBP];
05b3e0c2 1347#ifdef CONFIG_X86_64
6aa8b732
AK		 1348 regs->r8 = vcpu->regs[VCPU_REGS_R8];
1349 regs->r9 = vcpu->regs[VCPU_REGS_R9];
1350 regs->r10 = vcpu->regs[VCPU_REGS_R10];
1351 regs->r11 = vcpu->regs[VCPU_REGS_R11];
1352 regs->r12 = vcpu->regs[VCPU_REGS_R12];
1353 regs->r13 = vcpu->regs[VCPU_REGS_R13];
1354 regs->r14 = vcpu->regs[VCPU_REGS_R14];
1355 regs->r15 = vcpu->regs[VCPU_REGS_R15];
1356#endif
1357
1358 regs->rip = vcpu->rip;
1359 regs->rflags = kvm_arch_ops->get_rflags(vcpu);
1360
1361 /*
1362 * Don't leak debug flags in case they were set for guest debugging
1363 */
1364 if (vcpu->guest_debug.enabled && vcpu->guest_debug.singlestep)
1365 regs->rflags &= ~(X86_EFLAGS_TF | X86_EFLAGS_RF);
1366
1367 vcpu_put(vcpu);
1368
1369 return 0;
1370}
1371
1372static int kvm_dev_ioctl_set_regs(struct kvm *kvm, struct kvm_regs *regs)
1373{
1374 struct kvm_vcpu *vcpu;
1375
5aacf0ca 1376 if (!valid_vcpu(regs->vcpu))
6aa8b732
AK		 1377 return -EINVAL;
1378
1379 vcpu = vcpu_load(kvm, regs->vcpu);
1380 if (!vcpu)
1381 return -ENOENT;
1382
1383 vcpu->regs[VCPU_REGS_RAX] = regs->rax;
1384 vcpu->regs[VCPU_REGS_RBX] = regs->rbx;
1385 vcpu->regs[VCPU_REGS_RCX] = regs->rcx;
1386 vcpu->regs[VCPU_REGS_RDX] = regs->rdx;
1387 vcpu->regs[VCPU_REGS_RSI] = regs->rsi;
1388 vcpu->regs[VCPU_REGS_RDI] = regs->rdi;
1389 vcpu->regs[VCPU_REGS_RSP] = regs->rsp;
1390 vcpu->regs[VCPU_REGS_RBP] = regs->rbp;
05b3e0c2 1391#ifdef CONFIG_X86_64
6aa8b732
AK		 1392 vcpu->regs[VCPU_REGS_R8] = regs->r8;
1393 vcpu->regs[VCPU_REGS_R9] = regs->r9;
1394 vcpu->regs[VCPU_REGS_R10] = regs->r10;
1395 vcpu->regs[VCPU_REGS_R11] = regs->r11;
1396 vcpu->regs[VCPU_REGS_R12] = regs->r12;
1397 vcpu->regs[VCPU_REGS_R13] = regs->r13;
1398 vcpu->regs[VCPU_REGS_R14] = regs->r14;
1399 vcpu->regs[VCPU_REGS_R15] = regs->r15;
1400#endif
1401
1402 vcpu->rip = regs->rip;
1403 kvm_arch_ops->set_rflags(vcpu, regs->rflags);
1404
1405 kvm_arch_ops->decache_regs(vcpu);
1406
1407 vcpu_put(vcpu);
1408
1409 return 0;
1410}
1411
1412static void get_segment(struct kvm_vcpu *vcpu,
1413 struct kvm_segment *var, int seg)
1414{
1415 return kvm_arch_ops->get_segment(vcpu, var, seg);
1416}
1417
1418static int kvm_dev_ioctl_get_sregs(struct kvm *kvm, struct kvm_sregs *sregs)
1419{
1420 struct kvm_vcpu *vcpu;
1421 struct descriptor_table dt;
1422
5aacf0ca 1423 if (!valid_vcpu(sregs->vcpu))
6aa8b732
AK		 1424 return -EINVAL;
1425 vcpu = vcpu_load(kvm, sregs->vcpu);
1426 if (!vcpu)
1427 return -ENOENT;
1428
1429 get_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
1430 get_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
1431 get_segment(vcpu, &sregs->es, VCPU_SREG_ES);
1432 get_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
1433 get_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
1434 get_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
1435
1436 get_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
1437 get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
1438
1439 kvm_arch_ops->get_idt(vcpu, &dt);
1440 sregs->idt.limit = dt.limit;
1441 sregs->idt.base = dt.base;
1442 kvm_arch_ops->get_gdt(vcpu, &dt);
1443 sregs->gdt.limit = dt.limit;
1444 sregs->gdt.base = dt.base;
1445
399badf3 1446 kvm_arch_ops->decache_cr0_cr4_guest_bits(vcpu);
6aa8b732
AK		 1447 sregs->cr0 = vcpu->cr0;
1448 sregs->cr2 = vcpu->cr2;
1449 sregs->cr3 = vcpu->cr3;
1450 sregs->cr4 = vcpu->cr4;
1451 sregs->cr8 = vcpu->cr8;
1452 sregs->efer = vcpu->shadow_efer;
1453 sregs->apic_base = vcpu->apic_base;
1454
1455 memcpy(sregs->interrupt_bitmap, vcpu->irq_pending,
1456 sizeof sregs->interrupt_bitmap);
1457
1458 vcpu_put(vcpu);
1459
1460 return 0;
1461}
1462
1463static void set_segment(struct kvm_vcpu *vcpu,
1464 struct kvm_segment *var, int seg)
1465{
1466 return kvm_arch_ops->set_segment(vcpu, var, seg);
1467}
1468
1469static int kvm_dev_ioctl_set_sregs(struct kvm *kvm, struct kvm_sregs *sregs)
1470{
1471 struct kvm_vcpu *vcpu;
1472 int mmu_reset_needed = 0;
1473 int i;
1474 struct descriptor_table dt;
1475
5aacf0ca 1476 if (!valid_vcpu(sregs->vcpu))
6aa8b732
AK		 1477 return -EINVAL;
1478 vcpu = vcpu_load(kvm, sregs->vcpu);
1479 if (!vcpu)
1480 return -ENOENT;
1481
1482 set_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
1483 set_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
1484 set_segment(vcpu, &sregs->es, VCPU_SREG_ES);
1485 set_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
1486 set_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
1487 set_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
1488
1489 set_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
1490 set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
1491
1492 dt.limit = sregs->idt.limit;
1493 dt.base = sregs->idt.base;
1494 kvm_arch_ops->set_idt(vcpu, &dt);
1495 dt.limit = sregs->gdt.limit;
1496 dt.base = sregs->gdt.base;
1497 kvm_arch_ops->set_gdt(vcpu, &dt);
1498
1499 vcpu->cr2 = sregs->cr2;
1500 mmu_reset_needed |= vcpu->cr3 != sregs->cr3;
1501 vcpu->cr3 = sregs->cr3;
1502
1503 vcpu->cr8 = sregs->cr8;
1504
1505 mmu_reset_needed |= vcpu->shadow_efer != sregs->efer;
05b3e0c2 1506#ifdef CONFIG_X86_64
6aa8b732
AK		 1507 kvm_arch_ops->set_efer(vcpu, sregs->efer);
1508#endif
1509 vcpu->apic_base = sregs->apic_base;
1510
399badf3
AK		 1511 kvm_arch_ops->decache_cr0_cr4_guest_bits(vcpu);
1512
6aa8b732
AK		 1513 mmu_reset_needed |= vcpu->cr0 != sregs->cr0;
1514 kvm_arch_ops->set_cr0_no_modeswitch(vcpu, sregs->cr0);
1515
1516 mmu_reset_needed |= vcpu->cr4 != sregs->cr4;
1517 kvm_arch_ops->set_cr4(vcpu, sregs->cr4);
1b0973bd
AK		 1518 if (!is_long_mode(vcpu) && is_pae(vcpu))
1519 load_pdptrs(vcpu, vcpu->cr3);
6aa8b732
AK		 1520
1521 if (mmu_reset_needed)
1522 kvm_mmu_reset_context(vcpu);
1523
1524 memcpy(vcpu->irq_pending, sregs->interrupt_bitmap,
1525 sizeof vcpu->irq_pending);
1526 vcpu->irq_summary = 0;
1527 for (i = 0; i < NR_IRQ_WORDS; ++i)
1528 if (vcpu->irq_pending[i])
1529 __set_bit(i, &vcpu->irq_summary);
1530
1531 vcpu_put(vcpu);
1532
1533 return 0;
1534}
1535
1536/*
1537 * List of msr numbers which we expose to userspace through KVM_GET_MSRS
1538 * and KVM_SET_MSRS, and KVM_GET_MSR_INDEX_LIST.
bf591b24
MR		 1539 *
1540 * This list is modified at module load time to reflect the
1541 * capabilities of the host cpu.
6aa8b732
AK		 1542 */
1543static u32 msrs_to_save[] = {
1544 MSR_IA32_SYSENTER_CS, MSR_IA32_SYSENTER_ESP, MSR_IA32_SYSENTER_EIP,
1545 MSR_K6_STAR,
05b3e0c2 1546#ifdef CONFIG_X86_64
6aa8b732
AK		 1547 MSR_CSTAR, MSR_KERNEL_GS_BASE, MSR_SYSCALL_MASK, MSR_LSTAR,
1548#endif
1549 MSR_IA32_TIME_STAMP_COUNTER,
1550};
1551
bf591b24
MR		 1552static unsigned num_msrs_to_save;
1553
1554static __init void kvm_init_msr_list(void)
1555{
1556 u32 dummy[2];
1557 unsigned i, j;
1558
1559 for (i = j = 0; i < ARRAY_SIZE(msrs_to_save); i++) {
1560 if (rdmsr_safe(msrs_to_save[i], &dummy[0], &dummy[1]) < 0)
1561 continue;
1562 if (j < i)
1563 msrs_to_save[j] = msrs_to_save[i];
1564 j++;
1565 }
1566 num_msrs_to_save = j;
1567}
6aa8b732
AK		 1568
1569/*
1570 * Adapt set_msr() to msr_io()'s calling convention
1571 */
1572static int do_set_msr(struct kvm_vcpu *vcpu, unsigned index, u64 *data)
1573{
1574 return set_msr(vcpu, index, *data);
1575}
1576
1577/*
1578 * Read or write a bunch of msrs. All parameters are kernel addresses.
1579 *
1580 * @return number of msrs set successfully.
1581 */
1582static int __msr_io(struct kvm *kvm, struct kvm_msrs *msrs,
1583 struct kvm_msr_entry *entries,
1584 int (*do_msr)(struct kvm_vcpu *vcpu,
1585 unsigned index, u64 *data))
1586{
1587 struct kvm_vcpu *vcpu;
1588 int i;
1589
5aacf0ca 1590 if (!valid_vcpu(msrs->vcpu))
6aa8b732
AK		 1591 return -EINVAL;
1592
1593 vcpu = vcpu_load(kvm, msrs->vcpu);
1594 if (!vcpu)
1595 return -ENOENT;
1596
1597 for (i = 0; i < msrs->nmsrs; ++i)
1598 if (do_msr(vcpu, entries[i].index, &entries[i].data))
1599 break;
1600
1601 vcpu_put(vcpu);
1602
1603 return i;
1604}
1605
1606/*
1607 * Read or write a bunch of msrs. Parameters are user addresses.
1608 *
1609 * @return number of msrs set successfully.
1610 */
1611static int msr_io(struct kvm *kvm, struct kvm_msrs __user *user_msrs,
1612 int (*do_msr)(struct kvm_vcpu *vcpu,
1613 unsigned index, u64 *data),
1614 int writeback)
1615{
1616 struct kvm_msrs msrs;
1617 struct kvm_msr_entry *entries;
1618 int r, n;
1619 unsigned size;
1620
1621 r = -EFAULT;
1622 if (copy_from_user(&msrs, user_msrs, sizeof msrs))
1623 goto out;
1624
1625 r = -E2BIG;
1626 if (msrs.nmsrs >= MAX_IO_MSRS)
1627 goto out;
1628
1629 r = -ENOMEM;
1630 size = sizeof(struct kvm_msr_entry) * msrs.nmsrs;
1631 entries = vmalloc(size);
1632 if (!entries)
1633 goto out;
1634
1635 r = -EFAULT;
1636 if (copy_from_user(entries, user_msrs->entries, size))
1637 goto out_free;
1638
1639 r = n = __msr_io(kvm, &msrs, entries, do_msr);
1640 if (r < 0)
1641 goto out_free;
1642
1643 r = -EFAULT;
1644 if (writeback && copy_to_user(user_msrs->entries, entries, size))
1645 goto out_free;
1646
1647 r = n;
1648
1649out_free:
1650 vfree(entries);
1651out:
1652 return r;
1653}
1654
1655/*
1656 * Translate a guest virtual address to a guest physical address.
1657 */
1658static int kvm_dev_ioctl_translate(struct kvm *kvm, struct kvm_translation *tr)
1659{
1660 unsigned long vaddr = tr->linear_address;
1661 struct kvm_vcpu *vcpu;
1662 gpa_t gpa;
1663
1664 vcpu = vcpu_load(kvm, tr->vcpu);
1665 if (!vcpu)
1666 return -ENOENT;
1667 spin_lock(&kvm->lock);
1668 gpa = vcpu->mmu.gva_to_gpa(vcpu, vaddr);
1669 tr->physical_address = gpa;
1670 tr->valid = gpa != UNMAPPED_GVA;
1671 tr->writeable = 1;
1672 tr->usermode = 0;
1673 spin_unlock(&kvm->lock);
1674 vcpu_put(vcpu);
1675
1676 return 0;
1677}
1678
1679static int kvm_dev_ioctl_interrupt(struct kvm *kvm, struct kvm_interrupt *irq)
1680{
1681 struct kvm_vcpu *vcpu;
1682
5aacf0ca 1683 if (!valid_vcpu(irq->vcpu))
6aa8b732
AK		 1684 return -EINVAL;
1685 if (irq->irq < 0 || irq->irq >= 256)
1686 return -EINVAL;
1687 vcpu = vcpu_load(kvm, irq->vcpu);
1688 if (!vcpu)
1689 return -ENOENT;
1690
1691 set_bit(irq->irq, vcpu->irq_pending);
1692 set_bit(irq->irq / BITS_PER_LONG, &vcpu->irq_summary);
1693
1694 vcpu_put(vcpu);
1695
1696 return 0;
1697}
1698
1699static int kvm_dev_ioctl_debug_guest(struct kvm *kvm,
1700 struct kvm_debug_guest *dbg)
1701{
1702 struct kvm_vcpu *vcpu;
1703 int r;
1704
5aacf0ca 1705 if (!valid_vcpu(dbg->vcpu))
6aa8b732
AK		 1706 return -EINVAL;
1707 vcpu = vcpu_load(kvm, dbg->vcpu);
1708 if (!vcpu)
1709 return -ENOENT;
1710
1711 r = kvm_arch_ops->set_guest_debug(vcpu, dbg);
1712
1713 vcpu_put(vcpu);
1714
1715 return r;
1716}
1717
1718static long kvm_dev_ioctl(struct file *filp,
1719 unsigned int ioctl, unsigned long arg)
1720{
1721 struct kvm *kvm = filp->private_data;
1722 int r = -EINVAL;
1723
1724 switch (ioctl) {
0b76e20b
AK		 1725 case KVM_GET_API_VERSION:
1726 r = KVM_API_VERSION;
1727 break;
6aa8b732
AK		 1728 case KVM_CREATE_VCPU: {
1729 r = kvm_dev_ioctl_create_vcpu(kvm, arg);
1730 if (r)
1731 goto out;
1732 break;
1733 }
1734 case KVM_RUN: {
1735 struct kvm_run kvm_run;
1736
1737 r = -EFAULT;
1738 if (copy_from_user(&kvm_run, (void *)arg, sizeof kvm_run))
1739 goto out;
1740 r = kvm_dev_ioctl_run(kvm, &kvm_run);
c1150d8c 1741 if (r < 0 && r != -EINTR)
6aa8b732 1742 goto out;
c1150d8c
DL		 1743 if (copy_to_user((void *)arg, &kvm_run, sizeof kvm_run)) {
1744 r = -EFAULT;
6aa8b732 1745 goto out;
c1150d8c 1746 }
6aa8b732
AK		 1747 break;
1748 }
1749 case KVM_GET_REGS: {
1750 struct kvm_regs kvm_regs;
1751
1752 r = -EFAULT;
1753 if (copy_from_user(&kvm_regs, (void *)arg, sizeof kvm_regs))
1754 goto out;
1755 r = kvm_dev_ioctl_get_regs(kvm, &kvm_regs);
1756 if (r)
1757 goto out;
1758 r = -EFAULT;
1759 if (copy_to_user((void *)arg, &kvm_regs, sizeof kvm_regs))
1760 goto out;
1761 r = 0;
1762 break;
1763 }
1764 case KVM_SET_REGS: {
1765 struct kvm_regs kvm_regs;
1766
1767 r = -EFAULT;
1768 if (copy_from_user(&kvm_regs, (void *)arg, sizeof kvm_regs))
1769 goto out;
1770 r = kvm_dev_ioctl_set_regs(kvm, &kvm_regs);
1771 if (r)
1772 goto out;
1773 r = 0;
1774 break;
1775 }
1776 case KVM_GET_SREGS: {
1777 struct kvm_sregs kvm_sregs;
1778
1779 r = -EFAULT;
1780 if (copy_from_user(&kvm_sregs, (void *)arg, sizeof kvm_sregs))
1781 goto out;
1782 r = kvm_dev_ioctl_get_sregs(kvm, &kvm_sregs);
1783 if (r)
1784 goto out;
1785 r = -EFAULT;
1786 if (copy_to_user((void *)arg, &kvm_sregs, sizeof kvm_sregs))
1787 goto out;
1788 r = 0;
1789 break;
1790 }
1791 case KVM_SET_SREGS: {
1792 struct kvm_sregs kvm_sregs;
1793
1794 r = -EFAULT;
1795 if (copy_from_user(&kvm_sregs, (void *)arg, sizeof kvm_sregs))
1796 goto out;
1797 r = kvm_dev_ioctl_set_sregs(kvm, &kvm_sregs);
1798 if (r)
1799 goto out;
1800 r = 0;
1801 break;
1802 }
1803 case KVM_TRANSLATE: {
1804 struct kvm_translation tr;
1805
1806 r = -EFAULT;
1807 if (copy_from_user(&tr, (void *)arg, sizeof tr))
1808 goto out;
1809 r = kvm_dev_ioctl_translate(kvm, &tr);
1810 if (r)
1811 goto out;
1812 r = -EFAULT;
1813 if (copy_to_user((void *)arg, &tr, sizeof tr))
1814 goto out;
1815 r = 0;
1816 break;
1817 }
1818 case KVM_INTERRUPT: {
1819 struct kvm_interrupt irq;
1820
1821 r = -EFAULT;
1822 if (copy_from_user(&irq, (void *)arg, sizeof irq))
1823 goto out;
1824 r = kvm_dev_ioctl_interrupt(kvm, &irq);
1825 if (r)
1826 goto out;
1827 r = 0;
1828 break;
1829 }
1830 case KVM_DEBUG_GUEST: {
1831 struct kvm_debug_guest dbg;
1832
1833 r = -EFAULT;
1834 if (copy_from_user(&dbg, (void *)arg, sizeof dbg))
1835 goto out;
1836 r = kvm_dev_ioctl_debug_guest(kvm, &dbg);
1837 if (r)
1838 goto out;
1839 r = 0;
1840 break;
1841 }
1842 case KVM_SET_MEMORY_REGION: {
1843 struct kvm_memory_region kvm_mem;
1844
1845 r = -EFAULT;
1846 if (copy_from_user(&kvm_mem, (void *)arg, sizeof kvm_mem))
1847 goto out;
1848 r = kvm_dev_ioctl_set_memory_region(kvm, &kvm_mem);
1849 if (r)
1850 goto out;
1851 break;
1852 }
1853 case KVM_GET_DIRTY_LOG: {
1854 struct kvm_dirty_log log;
1855
1856 r = -EFAULT;
1857 if (copy_from_user(&log, (void *)arg, sizeof log))
1858 goto out;
1859 r = kvm_dev_ioctl_get_dirty_log(kvm, &log);
1860 if (r)
1861 goto out;
1862 break;
1863 }
1864 case KVM_GET_MSRS:
1865 r = msr_io(kvm, (void __user *)arg, get_msr, 1);
1866 break;
1867 case KVM_SET_MSRS:
1868 r = msr_io(kvm, (void __user *)arg, do_set_msr, 0);
1869 break;
1870 case KVM_GET_MSR_INDEX_LIST: {
1871 struct kvm_msr_list __user *user_msr_list = (void __user *)arg;
1872 struct kvm_msr_list msr_list;
1873 unsigned n;
1874
1875 r = -EFAULT;
1876 if (copy_from_user(&msr_list, user_msr_list, sizeof msr_list))
1877 goto out;
1878 n = msr_list.nmsrs;
bf591b24 1879 msr_list.nmsrs = num_msrs_to_save;
6aa8b732
AK		 1880 if (copy_to_user(user_msr_list, &msr_list, sizeof msr_list))
1881 goto out;
1882 r = -E2BIG;
bf591b24 1883 if (n < num_msrs_to_save)
6aa8b732
AK		 1884 goto out;
1885 r = -EFAULT;
1886 if (copy_to_user(user_msr_list->indices, &msrs_to_save,
bf591b24 1887 num_msrs_to_save * sizeof(u32)))
6aa8b732
AK		 1888 goto out;
1889 r = 0;
1890 }
1891 default:
1892 ;
1893 }
1894out:
1895 return r;
1896}
1897
1898static struct page *kvm_dev_nopage(struct vm_area_struct *vma,
1899 unsigned long address,
1900 int *type)
1901{
1902 struct kvm *kvm = vma->vm_file->private_data;
1903 unsigned long pgoff;
1904 struct kvm_memory_slot *slot;
1905 struct page *page;
1906
1907 *type = VM_FAULT_MINOR;
1908 pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
1909 slot = gfn_to_memslot(kvm, pgoff);
1910 if (!slot)
1911 return NOPAGE_SIGBUS;
1912 page = gfn_to_page(slot, pgoff);
1913 if (!page)
1914 return NOPAGE_SIGBUS;
1915 get_page(page);
1916 return page;
1917}
1918
1919static struct vm_operations_struct kvm_dev_vm_ops = {
1920 .nopage = kvm_dev_nopage,
1921};
1922
1923static int kvm_dev_mmap(struct file *file, struct vm_area_struct *vma)
1924{
1925 vma->vm_ops = &kvm_dev_vm_ops;
1926 return 0;
1927}
1928
1929static struct file_operations kvm_chardev_ops = {
1930 .open = kvm_dev_open,
1931 .release = kvm_dev_release,
1932 .unlocked_ioctl = kvm_dev_ioctl,
1933 .compat_ioctl = kvm_dev_ioctl,
1934 .mmap = kvm_dev_mmap,
1935};
1936
1937static struct miscdevice kvm_dev = {
1938 MISC_DYNAMIC_MINOR,
1939 "kvm",
1940 &kvm_chardev_ops,
1941};
1942
1943static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
1944 void *v)
1945{
1946 if (val == SYS_RESTART) {
1947 /*
1948 * Some (well, at least mine) BIOSes hang on reboot if
1949 * in vmx root mode.
1950 */
1951 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
1952 on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
1953 }
1954 return NOTIFY_OK;
1955}
1956
1957static struct notifier_block kvm_reboot_notifier = {
1958 .notifier_call = kvm_reboot,
1959 .priority = 0,
1960};
1961
1962static __init void kvm_init_debug(void)
1963{
1964 struct kvm_stats_debugfs_item *p;
1965
1966 debugfs_dir = debugfs_create_dir("kvm", 0);
1967 for (p = debugfs_entries; p->name; ++p)
1968 p->dentry = debugfs_create_u32(p->name, 0444, debugfs_dir,
1969 p->data);
1970}
1971
1972static void kvm_exit_debug(void)
1973{
1974 struct kvm_stats_debugfs_item *p;
1975
1976 for (p = debugfs_entries; p->name; ++p)
1977 debugfs_remove(p->dentry);
1978 debugfs_remove(debugfs_dir);
1979}
1980
1981hpa_t bad_page_address;
1982
1983int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module)
1984{
1985 int r;
1986
09db28b8
YI		 1987 if (kvm_arch_ops) {
1988 printk(KERN_ERR "kvm: already loaded the other module\n");
1989 return -EEXIST;
1990 }
1991
e097f35c 1992 if (!ops->cpu_has_kvm_support()) {
6aa8b732
AK		 1993 printk(KERN_ERR "kvm: no hardware support\n");
1994 return -EOPNOTSUPP;
1995 }
e097f35c 1996 if (ops->disabled_by_bios()) {
6aa8b732
AK		 1997 printk(KERN_ERR "kvm: disabled by bios\n");
1998 return -EOPNOTSUPP;
1999 }
2000
e097f35c
YI		 2001 kvm_arch_ops = ops;
2002
6aa8b732
AK		 2003 r = kvm_arch_ops->hardware_setup();
2004 if (r < 0)
2005 return r;
2006
2007 on_each_cpu(kvm_arch_ops->hardware_enable, 0, 0, 1);
2008 register_reboot_notifier(&kvm_reboot_notifier);
2009
2010 kvm_chardev_ops.owner = module;
2011
2012 r = misc_register(&kvm_dev);
2013 if (r) {
2014 printk (KERN_ERR "kvm: misc device register failed\n");
2015 goto out_free;
2016 }
2017
2018 return r;
2019
2020out_free:
2021 unregister_reboot_notifier(&kvm_reboot_notifier);
2022 on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
2023 kvm_arch_ops->hardware_unsetup();
2024 return r;
2025}
2026
2027void kvm_exit_arch(void)
2028{
2029 misc_deregister(&kvm_dev);
2030
2031 unregister_reboot_notifier(&kvm_reboot_notifier);
2032 on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
2033 kvm_arch_ops->hardware_unsetup();
09db28b8 2034 kvm_arch_ops = NULL;
6aa8b732
AK		 2035}
2036
2037static __init int kvm_init(void)
2038{
2039 static struct page *bad_page;
2040 int r = 0;
2041
2042 kvm_init_debug();
2043
bf591b24
MR		 2044 kvm_init_msr_list();
2045
6aa8b732
AK		 2046 if ((bad_page = alloc_page(GFP_KERNEL)) == NULL) {
2047 r = -ENOMEM;
2048 goto out;
2049 }
2050
2051 bad_page_address = page_to_pfn(bad_page) << PAGE_SHIFT;
2052 memset(__va(bad_page_address), 0, PAGE_SIZE);
2053
2054 return r;
2055
2056out:
2057 kvm_exit_debug();
2058 return r;
2059}
2060
2061static __exit void kvm_exit(void)
2062{
2063 kvm_exit_debug();
2064 __free_page(pfn_to_page(bad_page_address >> PAGE_SHIFT));
2065}
2066
2067module_init(kvm_init)
2068module_exit(kvm_exit)
2069
2070EXPORT_SYMBOL_GPL(kvm_init_arch);
2071EXPORT_SYMBOL_GPL(kvm_exit_arch);
Linux kernel - Deliverables
This page took 0.222722 seconds and 5 git commands to generate.