blob: d4eae6af07382dc404d278ff1bf67814f8449fb2 [file] [log] [blame]
Avi Kivity6aa8b732006-12-10 02:21:36 -08001/*
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
Hollis Blancharde2174022007-12-03 15:30:24 -060018#include "iodev.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080019
Avi Kivityedf88412007-12-16 11:02:48 +020020#include <linux/kvm_host.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080021#include <linux/kvm.h>
22#include <linux/module.h>
23#include <linux/errno.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080024#include <linux/percpu.h>
25#include <linux/gfp.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080026#include <linux/mm.h>
27#include <linux/miscdevice.h>
28#include <linux/vmalloc.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080029#include <linux/reboot.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080030#include <linux/debugfs.h>
31#include <linux/highmem.h>
32#include <linux/file.h>
Avi Kivity59ae6c62007-02-12 00:54:48 -080033#include <linux/sysdev.h>
Avi Kivity774c47f2007-02-12 00:54:47 -080034#include <linux/cpu.h>
Alexey Dobriyane8edc6e2007-05-21 01:22:52 +040035#include <linux/sched.h>
Avi Kivityd9e368d2007-06-07 19:18:30 +030036#include <linux/cpumask.h>
37#include <linux/smp.h>
Avi Kivityd6d28162007-06-28 08:38:16 -040038#include <linux/anon_inodes.h>
Avi Kivity04d2cc72007-09-10 18:10:54 +030039#include <linux/profile.h>
Anthony Liguori7aa81cc2007-09-17 14:57:50 -050040#include <linux/kvm_para.h>
Izik Eidus6fc138d2007-10-09 19:20:39 +020041#include <linux/pagemap.h>
Anthony Liguori8d4e1282007-10-18 09:59:34 -050042#include <linux/mman.h>
Anthony Liguori35149e22008-04-02 14:46:56 -050043#include <linux/swap.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080044
Avi Kivitye4956062007-06-28 14:15:57 -040045#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040046#include <asm/io.h>
47#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020048#include <asm/pgtable.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080049
50MODULE_AUTHOR("Qumranet");
51MODULE_LICENSE("GPL");
52
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080053DEFINE_SPINLOCK(kvm_lock);
54LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080055
Avi Kivity1b6c0162007-05-24 13:03:52 +030056static cpumask_t cpus_hardware_enabled;
57
Rusty Russellc16f8622007-07-30 21:12:19 +100058struct kmem_cache *kvm_vcpu_cache;
59EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030060
Avi Kivity15ad7142007-07-11 18:17:21 +030061static __read_mostly struct preempt_ops kvm_preempt_ops;
62
Hollis Blanchard76f7c872008-04-15 16:05:42 -050063struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080064
Avi Kivitybccf2152007-02-21 18:04:26 +020065static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
66 unsigned long arg);
67
James Morris5aacf0c2006-12-22 01:04:55 -080068static inline int valid_vcpu(int n)
69{
70 return likely(n >= 0 && n < KVM_MAX_VCPUS);
71}
72
Avi Kivity6aa8b732006-12-10 02:21:36 -080073/*
74 * Switches to specified vcpu, until a matching vcpu_put()
75 */
Carsten Otte313a3dc2007-10-11 19:16:52 +020076void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -080077{
Avi Kivity15ad7142007-07-11 18:17:21 +030078 int cpu;
79
Avi Kivitybccf2152007-02-21 18:04:26 +020080 mutex_lock(&vcpu->mutex);
Avi Kivity15ad7142007-07-11 18:17:21 +030081 cpu = get_cpu();
82 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +020083 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +030084 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +020085}
86
Carsten Otte313a3dc2007-10-11 19:16:52 +020087void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -080088{
Avi Kivity15ad7142007-07-11 18:17:21 +030089 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +020090 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +030091 preempt_notifier_unregister(&vcpu->preempt_notifier);
92 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -080093 mutex_unlock(&vcpu->mutex);
94}
95
Avi Kivityd9e368d2007-06-07 19:18:30 +030096static void ack_flush(void *_completed)
97{
Avi Kivityd9e368d2007-06-07 19:18:30 +030098}
99
100void kvm_flush_remote_tlbs(struct kvm *kvm)
101{
Laurent Vivier49d3bd72007-10-22 16:33:07 +0200102 int i, cpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300103 cpumask_t cpus;
104 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300105
Avi Kivityd9e368d2007-06-07 19:18:30 +0300106 cpus_clear(cpus);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000107 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
108 vcpu = kvm->vcpus[i];
109 if (!vcpu)
110 continue;
Avi Kivity3176bc32007-10-16 17:22:08 +0200111 if (test_and_set_bit(KVM_REQ_TLB_FLUSH, &vcpu->requests))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300112 continue;
113 cpu = vcpu->cpu;
114 if (cpu != -1 && cpu != raw_smp_processor_id())
Laurent Vivier49d3bd72007-10-22 16:33:07 +0200115 cpu_set(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300116 }
Avi Kivity0f74a242007-11-20 23:01:14 +0200117 if (cpus_empty(cpus))
118 return;
119 ++kvm->stat.remote_tlb_flush;
Laurent Vivier49d3bd72007-10-22 16:33:07 +0200120 smp_call_function_mask(cpus, ack_flush, NULL, 1);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300121}
122
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500123void kvm_reload_remote_mmus(struct kvm *kvm)
124{
125 int i, cpu;
126 cpumask_t cpus;
127 struct kvm_vcpu *vcpu;
128
129 cpus_clear(cpus);
130 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
131 vcpu = kvm->vcpus[i];
132 if (!vcpu)
133 continue;
134 if (test_and_set_bit(KVM_REQ_MMU_RELOAD, &vcpu->requests))
135 continue;
136 cpu = vcpu->cpu;
137 if (cpu != -1 && cpu != raw_smp_processor_id())
138 cpu_set(cpu, cpus);
139 }
140 if (cpus_empty(cpus))
141 return;
142 smp_call_function_mask(cpus, ack_flush, NULL, 1);
143}
144
145
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000146int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
147{
148 struct page *page;
149 int r;
150
151 mutex_init(&vcpu->mutex);
152 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000153 vcpu->kvm = kvm;
154 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300155 init_waitqueue_head(&vcpu->wq);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000156
157 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
158 if (!page) {
159 r = -ENOMEM;
160 goto fail;
161 }
162 vcpu->run = page_address(page);
163
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800164 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000165 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800166 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000167 return 0;
168
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000169fail_free_run:
170 free_page((unsigned long)vcpu->run);
171fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000172 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000173}
174EXPORT_SYMBOL_GPL(kvm_vcpu_init);
175
176void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
177{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800178 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000179 free_page((unsigned long)vcpu->run);
180}
181EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
182
Avi Kivityf17abe92007-02-21 19:28:04 +0200183static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800184{
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800185 struct kvm *kvm = kvm_arch_create_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800186
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800187 if (IS_ERR(kvm))
188 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800189
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200190 kvm->mm = current->mm;
191 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500192 spin_lock_init(&kvm->mmu_lock);
Eddie Dong74906342007-06-19 18:05:03 +0300193 kvm_io_bus_init(&kvm->pio_bus);
Shaohua Li11ec2802007-07-23 14:51:37 +0800194 mutex_init(&kvm->lock);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -0400195 kvm_io_bus_init(&kvm->mmio_bus);
Izik Eidus72dc67a2008-02-10 18:04:15 +0200196 init_rwsem(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300197 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000198 spin_lock(&kvm_lock);
199 list_add(&kvm->vm_list, &vm_list);
200 spin_unlock(&kvm_lock);
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800201out:
Avi Kivityf17abe92007-02-21 19:28:04 +0200202 return kvm;
203}
204
Avi Kivity6aa8b732006-12-10 02:21:36 -0800205/*
206 * Free any memory in @free but not in @dont.
207 */
208static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
209 struct kvm_memory_slot *dont)
210{
Izik Eidus290fc382007-09-27 14:11:22 +0200211 if (!dont || free->rmap != dont->rmap)
212 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800213
214 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
215 vfree(free->dirty_bitmap);
216
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300217 if (!dont || free->lpage_info != dont->lpage_info)
218 vfree(free->lpage_info);
219
Avi Kivity6aa8b732006-12-10 02:21:36 -0800220 free->npages = 0;
Al Viro8b6d44c2007-02-09 16:38:40 +0000221 free->dirty_bitmap = NULL;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500222 free->rmap = NULL;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300223 free->lpage_info = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800224}
225
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800226void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800227{
228 int i;
229
230 for (i = 0; i < kvm->nmemslots; ++i)
Al Viro8b6d44c2007-02-09 16:38:40 +0000231 kvm_free_physmem_slot(&kvm->memslots[i], NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800232}
233
Avi Kivityf17abe92007-02-21 19:28:04 +0200234static void kvm_destroy_vm(struct kvm *kvm)
235{
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200236 struct mm_struct *mm = kvm->mm;
237
Avi Kivity133de902007-02-12 00:54:44 -0800238 spin_lock(&kvm_lock);
239 list_del(&kvm->vm_list);
240 spin_unlock(&kvm_lock);
Eddie Dong74906342007-06-19 18:05:03 +0300241 kvm_io_bus_destroy(&kvm->pio_bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -0400242 kvm_io_bus_destroy(&kvm->mmio_bus);
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800243 kvm_arch_destroy_vm(kvm);
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200244 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200245}
246
Izik Eidusd39f13b2008-03-30 16:01:25 +0300247void kvm_get_kvm(struct kvm *kvm)
248{
249 atomic_inc(&kvm->users_count);
250}
251EXPORT_SYMBOL_GPL(kvm_get_kvm);
252
253void kvm_put_kvm(struct kvm *kvm)
254{
255 if (atomic_dec_and_test(&kvm->users_count))
256 kvm_destroy_vm(kvm);
257}
258EXPORT_SYMBOL_GPL(kvm_put_kvm);
259
260
Avi Kivityf17abe92007-02-21 19:28:04 +0200261static int kvm_vm_release(struct inode *inode, struct file *filp)
262{
263 struct kvm *kvm = filp->private_data;
264
Izik Eidusd39f13b2008-03-30 16:01:25 +0300265 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800266 return 0;
267}
268
Avi Kivity6aa8b732006-12-10 02:21:36 -0800269/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800270 * Allocate some memory and give it an address in the guest physical address
271 * space.
272 *
273 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800274 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500275 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800276 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800277int __kvm_set_memory_region(struct kvm *kvm,
278 struct kvm_userspace_memory_region *mem,
279 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800280{
281 int r;
282 gfn_t base_gfn;
283 unsigned long npages;
284 unsigned long i;
285 struct kvm_memory_slot *memslot;
286 struct kvm_memory_slot old, new;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800287
288 r = -EINVAL;
289 /* General sanity checks */
290 if (mem->memory_size & (PAGE_SIZE - 1))
291 goto out;
292 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
293 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200294 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800295 goto out;
296 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
297 goto out;
298
299 memslot = &kvm->memslots[mem->slot];
300 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
301 npages = mem->memory_size >> PAGE_SHIFT;
302
303 if (!npages)
304 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
305
Avi Kivity6aa8b732006-12-10 02:21:36 -0800306 new = old = *memslot;
307
308 new.base_gfn = base_gfn;
309 new.npages = npages;
310 new.flags = mem->flags;
311
312 /* Disallow changing a memory slot's size. */
313 r = -EINVAL;
314 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800315 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800316
317 /* Check for overlaps */
318 r = -EEXIST;
319 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
320 struct kvm_memory_slot *s = &kvm->memslots[i];
321
322 if (s == memslot)
323 continue;
324 if (!((base_gfn + npages <= s->base_gfn) ||
325 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800326 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800327 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800328
Avi Kivity6aa8b732006-12-10 02:21:36 -0800329 /* Free page dirty bitmap if unneeded */
330 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000331 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800332
333 r = -ENOMEM;
334
335 /* Allocate if a slot is being created */
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500336 if (npages && !new.rmap) {
Mike Dayd77c26f2007-10-08 09:02:08 -0400337 new.rmap = vmalloc(npages * sizeof(struct page *));
Izik Eidus290fc382007-09-27 14:11:22 +0200338
339 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800340 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200341
Izik Eidus290fc382007-09-27 14:11:22 +0200342 memset(new.rmap, 0, npages * sizeof(*new.rmap));
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500343
Izik Eidus80b14b52007-10-25 11:54:04 +0200344 new.user_alloc = user_alloc;
Zhang Xiantao0de10342007-11-20 16:25:04 +0800345 new.userspace_addr = mem->userspace_addr;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800346 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300347 if (npages && !new.lpage_info) {
348 int largepages = npages / KVM_PAGES_PER_HPAGE;
349 if (npages % KVM_PAGES_PER_HPAGE)
350 largepages++;
351 if (base_gfn % KVM_PAGES_PER_HPAGE)
352 largepages++;
353
354 new.lpage_info = vmalloc(largepages * sizeof(*new.lpage_info));
355
356 if (!new.lpage_info)
357 goto out_free;
358
359 memset(new.lpage_info, 0, largepages * sizeof(*new.lpage_info));
360
361 if (base_gfn % KVM_PAGES_PER_HPAGE)
362 new.lpage_info[0].write_count = 1;
363 if ((base_gfn+npages) % KVM_PAGES_PER_HPAGE)
364 new.lpage_info[largepages-1].write_count = 1;
365 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800366
367 /* Allocate page dirty bitmap if needed */
368 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
369 unsigned dirty_bytes = ALIGN(npages, BITS_PER_LONG) / 8;
370
371 new.dirty_bitmap = vmalloc(dirty_bytes);
372 if (!new.dirty_bitmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800373 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800374 memset(new.dirty_bitmap, 0, dirty_bytes);
375 }
376
Avi Kivity6aa8b732006-12-10 02:21:36 -0800377 if (mem->slot >= kvm->nmemslots)
378 kvm->nmemslots = mem->slot + 1;
379
380 *memslot = new;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800381
Zhang Xiantao0de10342007-11-20 16:25:04 +0800382 r = kvm_arch_set_memory_region(kvm, mem, old, user_alloc);
383 if (r) {
384 *memslot = old;
385 goto out_free;
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800386 }
387
Avi Kivity6aa8b732006-12-10 02:21:36 -0800388 kvm_free_physmem_slot(&old, &new);
389 return 0;
390
Sheng Yangf78e0e22007-10-29 09:40:42 +0800391out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800392 kvm_free_physmem_slot(&new, &old);
393out:
394 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200395
396}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800397EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
398
399int kvm_set_memory_region(struct kvm *kvm,
400 struct kvm_userspace_memory_region *mem,
401 int user_alloc)
402{
403 int r;
404
Izik Eidus72dc67a2008-02-10 18:04:15 +0200405 down_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800406 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Izik Eidus72dc67a2008-02-10 18:04:15 +0200407 up_write(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800408 return r;
409}
Izik Eidus210c7c42007-10-24 23:52:57 +0200410EXPORT_SYMBOL_GPL(kvm_set_memory_region);
411
Carsten Otte1fe779f2007-10-29 16:08:35 +0100412int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
413 struct
414 kvm_userspace_memory_region *mem,
415 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200416{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200417 if (mem->slot >= KVM_MEMORY_SLOTS)
418 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200419 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800420}
421
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800422int kvm_get_dirty_log(struct kvm *kvm,
423 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800424{
425 struct kvm_memory_slot *memslot;
426 int r, i;
427 int n;
428 unsigned long any = 0;
429
Avi Kivity6aa8b732006-12-10 02:21:36 -0800430 r = -EINVAL;
431 if (log->slot >= KVM_MEMORY_SLOTS)
432 goto out;
433
434 memslot = &kvm->memslots[log->slot];
435 r = -ENOENT;
436 if (!memslot->dirty_bitmap)
437 goto out;
438
Uri Lublincd1a4a92007-02-22 16:43:09 +0200439 n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800440
Uri Lublincd1a4a92007-02-22 16:43:09 +0200441 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800442 any = memslot->dirty_bitmap[i];
443
444 r = -EFAULT;
445 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
446 goto out;
447
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800448 if (any)
449 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800450
451 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800452out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800453 return r;
454}
455
Izik Eiduscea7bb22007-10-17 19:17:48 +0200456int is_error_page(struct page *page)
457{
458 return page == bad_page;
459}
460EXPORT_SYMBOL_GPL(is_error_page);
461
Anthony Liguori35149e22008-04-02 14:46:56 -0500462int is_error_pfn(pfn_t pfn)
463{
464 return pfn == bad_pfn;
465}
466EXPORT_SYMBOL_GPL(is_error_pfn);
467
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200468static inline unsigned long bad_hva(void)
469{
470 return PAGE_OFFSET;
471}
472
473int kvm_is_error_hva(unsigned long addr)
474{
475 return addr == bad_hva();
476}
477EXPORT_SYMBOL_GPL(kvm_is_error_hva);
478
Avi Kivitye8207542007-03-30 16:54:30 +0300479static struct kvm_memory_slot *__gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800480{
481 int i;
482
483 for (i = 0; i < kvm->nmemslots; ++i) {
484 struct kvm_memory_slot *memslot = &kvm->memslots[i];
485
486 if (gfn >= memslot->base_gfn
487 && gfn < memslot->base_gfn + memslot->npages)
488 return memslot;
489 }
Al Viro8b6d44c2007-02-09 16:38:40 +0000490 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800491}
Avi Kivitye8207542007-03-30 16:54:30 +0300492
493struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
494{
495 gfn = unalias_gfn(kvm, gfn);
496 return __gfn_to_memslot(kvm, gfn);
497}
Avi Kivity6aa8b732006-12-10 02:21:36 -0800498
Izik Eiduse0d62c72007-10-24 23:57:46 +0200499int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
500{
501 int i;
502
503 gfn = unalias_gfn(kvm, gfn);
504 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
505 struct kvm_memory_slot *memslot = &kvm->memslots[i];
506
507 if (gfn >= memslot->base_gfn
508 && gfn < memslot->base_gfn + memslot->npages)
509 return 1;
510 }
511 return 0;
512}
513EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
514
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300515unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
Izik Eidus539cb662007-11-11 22:05:04 +0200516{
517 struct kvm_memory_slot *slot;
518
519 gfn = unalias_gfn(kvm, gfn);
520 slot = __gfn_to_memslot(kvm, gfn);
521 if (!slot)
522 return bad_hva();
523 return (slot->userspace_addr + (gfn - slot->base_gfn) * PAGE_SIZE);
524}
Sheng Yang0d150292008-04-25 21:44:50 +0800525EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +0200526
Anthony Liguoriaab61cc2007-10-29 15:15:20 -0500527/*
528 * Requires current->mm->mmap_sem to be held
529 */
Anthony Liguori35149e22008-04-02 14:46:56 -0500530pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
Avi Kivity954bbbc22007-03-30 14:02:32 +0300531{
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500532 struct page *page[1];
Izik Eidus539cb662007-11-11 22:05:04 +0200533 unsigned long addr;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500534 int npages;
Avi Kivity954bbbc22007-03-30 14:02:32 +0300535
Avi Kivity60395222007-10-21 11:03:36 +0200536 might_sleep();
537
Izik Eidus539cb662007-11-11 22:05:04 +0200538 addr = gfn_to_hva(kvm, gfn);
539 if (kvm_is_error_hva(addr)) {
Izik Eidus8a7ae052007-10-18 11:09:33 +0200540 get_page(bad_page);
Anthony Liguori35149e22008-04-02 14:46:56 -0500541 return page_to_pfn(bad_page);
Izik Eidus8a7ae052007-10-18 11:09:33 +0200542 }
Izik Eidus8a7ae052007-10-18 11:09:33 +0200543
Izik Eidus539cb662007-11-11 22:05:04 +0200544 npages = get_user_pages(current, current->mm, addr, 1, 1, 1, page,
545 NULL);
546
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500547 if (npages != 1) {
548 get_page(bad_page);
Anthony Liguori35149e22008-04-02 14:46:56 -0500549 return page_to_pfn(bad_page);
Izik Eidus8a7ae052007-10-18 11:09:33 +0200550 }
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500551
Anthony Liguori35149e22008-04-02 14:46:56 -0500552 return page_to_pfn(page[0]);
553}
554
555EXPORT_SYMBOL_GPL(gfn_to_pfn);
556
557struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
558{
559 return pfn_to_page(gfn_to_pfn(kvm, gfn));
Avi Kivity954bbbc22007-03-30 14:02:32 +0300560}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -0500561
Avi Kivity954bbbc22007-03-30 14:02:32 +0300562EXPORT_SYMBOL_GPL(gfn_to_page);
563
Izik Eidusb4231d62007-11-20 11:49:33 +0200564void kvm_release_page_clean(struct page *page)
565{
Anthony Liguori35149e22008-04-02 14:46:56 -0500566 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +0200567}
568EXPORT_SYMBOL_GPL(kvm_release_page_clean);
569
Anthony Liguori35149e22008-04-02 14:46:56 -0500570void kvm_release_pfn_clean(pfn_t pfn)
571{
572 put_page(pfn_to_page(pfn));
573}
574EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
575
Izik Eidusb4231d62007-11-20 11:49:33 +0200576void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +0200577{
Anthony Liguori35149e22008-04-02 14:46:56 -0500578 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +0200579}
Izik Eidusb4231d62007-11-20 11:49:33 +0200580EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +0200581
Anthony Liguori35149e22008-04-02 14:46:56 -0500582void kvm_release_pfn_dirty(pfn_t pfn)
583{
584 kvm_set_pfn_dirty(pfn);
585 kvm_release_pfn_clean(pfn);
586}
587EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
588
589void kvm_set_page_dirty(struct page *page)
590{
591 kvm_set_pfn_dirty(page_to_pfn(page));
592}
593EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
594
595void kvm_set_pfn_dirty(pfn_t pfn)
596{
597 struct page *page = pfn_to_page(pfn);
598 if (!PageReserved(page))
599 SetPageDirty(page);
600}
601EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
602
603void kvm_set_pfn_accessed(pfn_t pfn)
604{
605 mark_page_accessed(pfn_to_page(pfn));
606}
607EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
608
609void kvm_get_pfn(pfn_t pfn)
610{
611 get_page(pfn_to_page(pfn));
612}
613EXPORT_SYMBOL_GPL(kvm_get_pfn);
614
Izik Eidus195aefd2007-10-01 22:14:18 +0200615static int next_segment(unsigned long len, int offset)
616{
617 if (len > PAGE_SIZE - offset)
618 return PAGE_SIZE - offset;
619 else
620 return len;
621}
622
623int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
624 int len)
625{
Izik Eiduse0506bc2007-11-11 22:10:22 +0200626 int r;
627 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +0200628
Izik Eiduse0506bc2007-11-11 22:10:22 +0200629 addr = gfn_to_hva(kvm, gfn);
630 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +0200631 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +0200632 r = copy_from_user(data, (void __user *)addr + offset, len);
633 if (r)
634 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +0200635 return 0;
636}
637EXPORT_SYMBOL_GPL(kvm_read_guest_page);
638
639int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
640{
641 gfn_t gfn = gpa >> PAGE_SHIFT;
642 int seg;
643 int offset = offset_in_page(gpa);
644 int ret;
645
646 while ((seg = next_segment(len, offset)) != 0) {
647 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
648 if (ret < 0)
649 return ret;
650 offset = 0;
651 len -= seg;
652 data += seg;
653 ++gfn;
654 }
655 return 0;
656}
657EXPORT_SYMBOL_GPL(kvm_read_guest);
658
Marcelo Tosatti7ec54582007-12-20 19:18:23 -0500659int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
660 unsigned long len)
661{
662 int r;
663 unsigned long addr;
664 gfn_t gfn = gpa >> PAGE_SHIFT;
665 int offset = offset_in_page(gpa);
666
667 addr = gfn_to_hva(kvm, gfn);
668 if (kvm_is_error_hva(addr))
669 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +0100670 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -0500671 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +0100672 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -0500673 if (r)
674 return -EFAULT;
675 return 0;
676}
677EXPORT_SYMBOL(kvm_read_guest_atomic);
678
Izik Eidus195aefd2007-10-01 22:14:18 +0200679int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
680 int offset, int len)
681{
Izik Eiduse0506bc2007-11-11 22:10:22 +0200682 int r;
683 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +0200684
Izik Eiduse0506bc2007-11-11 22:10:22 +0200685 addr = gfn_to_hva(kvm, gfn);
686 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +0200687 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +0200688 r = copy_to_user((void __user *)addr + offset, data, len);
689 if (r)
690 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +0200691 mark_page_dirty(kvm, gfn);
692 return 0;
693}
694EXPORT_SYMBOL_GPL(kvm_write_guest_page);
695
696int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
697 unsigned long len)
698{
699 gfn_t gfn = gpa >> PAGE_SHIFT;
700 int seg;
701 int offset = offset_in_page(gpa);
702 int ret;
703
704 while ((seg = next_segment(len, offset)) != 0) {
705 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
706 if (ret < 0)
707 return ret;
708 offset = 0;
709 len -= seg;
710 data += seg;
711 ++gfn;
712 }
713 return 0;
714}
715
716int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
717{
Izik Eidus3e021bf2007-11-19 11:16:57 +0200718 return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +0200719}
720EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
721
722int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
723{
724 gfn_t gfn = gpa >> PAGE_SHIFT;
725 int seg;
726 int offset = offset_in_page(gpa);
727 int ret;
728
729 while ((seg = next_segment(len, offset)) != 0) {
730 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
731 if (ret < 0)
732 return ret;
733 offset = 0;
734 len -= seg;
735 ++gfn;
736 }
737 return 0;
738}
739EXPORT_SYMBOL_GPL(kvm_clear_guest);
740
Avi Kivity6aa8b732006-12-10 02:21:36 -0800741void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
742{
Nguyen Anh Quynh31389942007-06-05 10:35:19 +0300743 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800744
Uri Lublin3b6fff12007-10-30 10:42:09 +0200745 gfn = unalias_gfn(kvm, gfn);
Rusty Russell7e9d6192007-07-31 20:41:14 +1000746 memslot = __gfn_to_memslot(kvm, gfn);
747 if (memslot && memslot->dirty_bitmap) {
748 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800749
Rusty Russell7e9d6192007-07-31 20:41:14 +1000750 /* avoid RMW */
751 if (!test_bit(rel_gfn, memslot->dirty_bitmap))
752 set_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800753 }
754}
755
Eddie Dongb6958ce2007-07-18 12:15:21 +0300756/*
757 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
758 */
Hollis Blanchard8776e512007-10-31 17:24:24 -0500759void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +0300760{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -0300761 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +0300762
Marcelo Tosattie5c239c2008-05-08 19:47:01 -0300763 for (;;) {
764 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +0300765
Marcelo Tosattie5c239c2008-05-08 19:47:01 -0300766 if (kvm_cpu_has_interrupt(vcpu))
767 break;
768 if (kvm_cpu_has_pending_timer(vcpu))
769 break;
770 if (kvm_arch_vcpu_runnable(vcpu))
771 break;
772 if (signal_pending(current))
773 break;
774
Eddie Dongb6958ce2007-07-18 12:15:21 +0300775 vcpu_put(vcpu);
776 schedule();
777 vcpu_load(vcpu);
778 }
779
Marcelo Tosattie5c239c2008-05-08 19:47:01 -0300780 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +0300781}
782
Avi Kivity6aa8b732006-12-10 02:21:36 -0800783void kvm_resched(struct kvm_vcpu *vcpu)
784{
Yaozu Dong3fca0362007-04-25 16:49:19 +0300785 if (!need_resched())
786 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800787 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800788}
789EXPORT_SYMBOL_GPL(kvm_resched);
790
npiggin@suse.dee4a533a2007-12-05 18:15:52 +1100791static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200792{
793 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200794 struct page *page;
795
npiggin@suse.dee4a533a2007-12-05 18:15:52 +1100796 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +0200797 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +0200798#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +1100799 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +0800800 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +0200801#endif
Avi Kivity039576c2007-03-20 12:46:50 +0200802 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +1100803 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200804 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +1100805 vmf->page = page;
806 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200807}
808
809static struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +1100810 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200811};
812
813static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
814{
815 vma->vm_ops = &kvm_vcpu_vm_ops;
816 return 0;
817}
818
Avi Kivitybccf2152007-02-21 18:04:26 +0200819static int kvm_vcpu_release(struct inode *inode, struct file *filp)
820{
821 struct kvm_vcpu *vcpu = filp->private_data;
822
Al Viro66c0b392008-04-19 20:33:56 +0100823 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +0200824 return 0;
825}
826
Jan Engelhardt5c502742008-01-22 20:46:14 +0100827static const struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +0200828 .release = kvm_vcpu_release,
829 .unlocked_ioctl = kvm_vcpu_ioctl,
830 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200831 .mmap = kvm_vcpu_mmap,
Avi Kivitybccf2152007-02-21 18:04:26 +0200832};
833
834/*
835 * Allocates an inode for the vcpu.
836 */
837static int create_vcpu_fd(struct kvm_vcpu *vcpu)
838{
Al Viro2030a422008-02-23 06:46:49 -0500839 int fd = anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu);
840 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +0100841 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +0200842 return fd;
Avi Kivitybccf2152007-02-21 18:04:26 +0200843}
844
Avi Kivityc5ea7662007-02-20 18:41:05 +0200845/*
846 * Creates some virtual cpus. Good luck creating more than one.
847 */
848static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n)
849{
850 int r;
851 struct kvm_vcpu *vcpu;
852
Avi Kivityc5ea7662007-02-20 18:41:05 +0200853 if (!valid_vcpu(n))
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000854 return -EINVAL;
Avi Kivityc5ea7662007-02-20 18:41:05 +0200855
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800856 vcpu = kvm_arch_vcpu_create(kvm, n);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000857 if (IS_ERR(vcpu))
858 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +0200859
Avi Kivity15ad7142007-07-11 18:17:21 +0300860 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
861
Avi Kivity26e52152007-11-20 15:30:24 +0200862 r = kvm_arch_vcpu_setup(vcpu);
863 if (r)
864 goto vcpu_destroy;
865
Shaohua Li11ec2802007-07-23 14:51:37 +0800866 mutex_lock(&kvm->lock);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000867 if (kvm->vcpus[n]) {
868 r = -EEXIST;
Shaohua Li11ec2802007-07-23 14:51:37 +0800869 mutex_unlock(&kvm->lock);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800870 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000871 }
872 kvm->vcpus[n] = vcpu;
Shaohua Li11ec2802007-07-23 14:51:37 +0800873 mutex_unlock(&kvm->lock);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000874
875 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +0100876 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +0200877 r = create_vcpu_fd(vcpu);
878 if (r < 0)
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000879 goto unlink;
Avi Kivitybccf2152007-02-21 18:04:26 +0200880 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +0200881
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000882unlink:
Shaohua Li11ec2802007-07-23 14:51:37 +0800883 mutex_lock(&kvm->lock);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000884 kvm->vcpus[n] = NULL;
Shaohua Li11ec2802007-07-23 14:51:37 +0800885 mutex_unlock(&kvm->lock);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800886vcpu_destroy:
Hollis Blanchardd40ccc62007-11-19 14:04:43 -0600887 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +0200888 return r;
889}
890
Avi Kivity1961d272007-03-05 19:46:05 +0200891static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
892{
893 if (sigset) {
894 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
895 vcpu->sigset_active = 1;
896 vcpu->sigset = *sigset;
897 } else
898 vcpu->sigset_active = 0;
899 return 0;
900}
901
Avi Kivitybccf2152007-02-21 18:04:26 +0200902static long kvm_vcpu_ioctl(struct file *filp,
903 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800904{
Avi Kivitybccf2152007-02-21 18:04:26 +0200905 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +0000906 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +0200907 int r;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800908
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200909 if (vcpu->kvm->mm != current->mm)
910 return -EIO;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800911 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +0200912 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +0200913 r = -EINVAL;
914 if (arg)
915 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -0500916 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800917 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800918 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800919 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800920
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800921 r = -ENOMEM;
922 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
923 if (!kvm_regs)
924 goto out;
925 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800926 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800927 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800928 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800929 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
930 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800931 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800932out_free1:
933 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800934 break;
935 }
936 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800937 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800938
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800939 r = -ENOMEM;
940 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
941 if (!kvm_regs)
942 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800943 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800944 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
945 goto out_free2;
946 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800947 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800948 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800949 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +0800950out_free2:
951 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800952 break;
953 }
954 case KVM_GET_SREGS: {
955 struct kvm_sregs kvm_sregs;
956
Avi Kivitybccf2152007-02-21 18:04:26 +0200957 memset(&kvm_sregs, 0, sizeof kvm_sregs);
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -0500958 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, &kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800959 if (r)
960 goto out;
961 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +0000962 if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs))
Avi Kivity6aa8b732006-12-10 02:21:36 -0800963 goto out;
964 r = 0;
965 break;
966 }
967 case KVM_SET_SREGS: {
968 struct kvm_sregs kvm_sregs;
969
970 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +0000971 if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs))
Avi Kivity6aa8b732006-12-10 02:21:36 -0800972 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -0500973 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, &kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800974 if (r)
975 goto out;
976 r = 0;
977 break;
978 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -0300979 case KVM_GET_MP_STATE: {
980 struct kvm_mp_state mp_state;
981
982 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
983 if (r)
984 goto out;
985 r = -EFAULT;
986 if (copy_to_user(argp, &mp_state, sizeof mp_state))
987 goto out;
988 r = 0;
989 break;
990 }
991 case KVM_SET_MP_STATE: {
992 struct kvm_mp_state mp_state;
993
994 r = -EFAULT;
995 if (copy_from_user(&mp_state, argp, sizeof mp_state))
996 goto out;
997 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
998 if (r)
999 goto out;
1000 r = 0;
1001 break;
1002 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001003 case KVM_TRANSLATE: {
1004 struct kvm_translation tr;
1005
1006 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001007 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001008 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001009 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001010 if (r)
1011 goto out;
1012 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001013 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001014 goto out;
1015 r = 0;
1016 break;
1017 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001018 case KVM_DEBUG_GUEST: {
1019 struct kvm_debug_guest dbg;
1020
1021 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001022 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001023 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001024 r = kvm_arch_vcpu_ioctl_debug_guest(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001025 if (r)
1026 goto out;
1027 r = 0;
1028 break;
1029 }
Avi Kivity1961d272007-03-05 19:46:05 +02001030 case KVM_SET_SIGNAL_MASK: {
1031 struct kvm_signal_mask __user *sigmask_arg = argp;
1032 struct kvm_signal_mask kvm_sigmask;
1033 sigset_t sigset, *p;
1034
1035 p = NULL;
1036 if (argp) {
1037 r = -EFAULT;
1038 if (copy_from_user(&kvm_sigmask, argp,
1039 sizeof kvm_sigmask))
1040 goto out;
1041 r = -EINVAL;
1042 if (kvm_sigmask.len != sizeof sigset)
1043 goto out;
1044 r = -EFAULT;
1045 if (copy_from_user(&sigset, sigmask_arg->sigset,
1046 sizeof sigset))
1047 goto out;
1048 p = &sigset;
1049 }
1050 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
1051 break;
1052 }
Avi Kivityb8836732007-04-01 16:34:31 +03001053 case KVM_GET_FPU: {
1054 struct kvm_fpu fpu;
1055
1056 memset(&fpu, 0, sizeof fpu);
Hollis Blanchardd0752062007-10-31 17:24:25 -05001057 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, &fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001058 if (r)
1059 goto out;
1060 r = -EFAULT;
1061 if (copy_to_user(argp, &fpu, sizeof fpu))
1062 goto out;
1063 r = 0;
1064 break;
1065 }
1066 case KVM_SET_FPU: {
1067 struct kvm_fpu fpu;
1068
1069 r = -EFAULT;
1070 if (copy_from_user(&fpu, argp, sizeof fpu))
1071 goto out;
Hollis Blanchardd0752062007-10-31 17:24:25 -05001072 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, &fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001073 if (r)
1074 goto out;
1075 r = 0;
1076 break;
1077 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001078 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001079 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001080 }
1081out:
1082 return r;
1083}
1084
1085static long kvm_vm_ioctl(struct file *filp,
1086 unsigned int ioctl, unsigned long arg)
1087{
1088 struct kvm *kvm = filp->private_data;
1089 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001090 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001091
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001092 if (kvm->mm != current->mm)
1093 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001094 switch (ioctl) {
1095 case KVM_CREATE_VCPU:
1096 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1097 if (r < 0)
1098 goto out;
1099 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001100 case KVM_SET_USER_MEMORY_REGION: {
1101 struct kvm_userspace_memory_region kvm_userspace_mem;
1102
1103 r = -EFAULT;
1104 if (copy_from_user(&kvm_userspace_mem, argp,
1105 sizeof kvm_userspace_mem))
1106 goto out;
1107
1108 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001109 if (r)
1110 goto out;
1111 break;
1112 }
1113 case KVM_GET_DIRTY_LOG: {
1114 struct kvm_dirty_log log;
1115
1116 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001117 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001118 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02001119 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001120 if (r)
1121 goto out;
1122 break;
1123 }
Avi Kivityf17abe92007-02-21 19:28:04 +02001124 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01001125 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02001126 }
1127out:
1128 return r;
1129}
1130
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001131static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02001132{
1133 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02001134 struct page *page;
1135
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001136 if (!kvm_is_visible_gfn(kvm, vmf->pgoff))
1137 return VM_FAULT_SIGBUS;
Marcelo Tosatti10589a42007-12-20 19:18:22 -05001138 page = gfn_to_page(kvm, vmf->pgoff);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001139 if (is_error_page(page)) {
Izik Eidusb4231d62007-11-20 11:49:33 +02001140 kvm_release_page_clean(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001141 return VM_FAULT_SIGBUS;
Izik Eidus8a7ae052007-10-18 11:09:33 +02001142 }
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001143 vmf->page = page;
1144 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02001145}
1146
1147static struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001148 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02001149};
1150
1151static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
1152{
1153 vma->vm_ops = &kvm_vm_vm_ops;
1154 return 0;
1155}
1156
Jan Engelhardt5c502742008-01-22 20:46:14 +01001157static const struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02001158 .release = kvm_vm_release,
1159 .unlocked_ioctl = kvm_vm_ioctl,
1160 .compat_ioctl = kvm_vm_ioctl,
1161 .mmap = kvm_vm_mmap,
1162};
1163
1164static int kvm_dev_ioctl_create_vm(void)
1165{
Al Viro2030a422008-02-23 06:46:49 -05001166 int fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001167 struct kvm *kvm;
1168
Avi Kivityf17abe92007-02-21 19:28:04 +02001169 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04001170 if (IS_ERR(kvm))
1171 return PTR_ERR(kvm);
Al Viro2030a422008-02-23 06:46:49 -05001172 fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm);
1173 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01001174 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02001175
Avi Kivityf17abe92007-02-21 19:28:04 +02001176 return fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001177}
1178
1179static long kvm_dev_ioctl(struct file *filp,
1180 unsigned int ioctl, unsigned long arg)
1181{
1182 void __user *argp = (void __user *)arg;
Avi Kivity07c45a32007-03-07 13:05:38 +02001183 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02001184
1185 switch (ioctl) {
1186 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001187 r = -EINVAL;
1188 if (arg)
1189 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02001190 r = KVM_API_VERSION;
1191 break;
1192 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001193 r = -EINVAL;
1194 if (arg)
1195 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02001196 r = kvm_dev_ioctl_create_vm();
1197 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08001198 case KVM_CHECK_EXTENSION:
1199 r = kvm_dev_ioctl_check_extension((long)argp);
Avi Kivity5d308f42007-03-01 17:56:20 +02001200 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02001201 case KVM_GET_VCPU_MMAP_SIZE:
1202 r = -EINVAL;
1203 if (arg)
1204 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02001205 r = PAGE_SIZE; /* struct kvm_run */
1206#ifdef CONFIG_X86
1207 r += PAGE_SIZE; /* pio data page */
1208#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02001209 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04001210 case KVM_TRACE_ENABLE:
1211 case KVM_TRACE_PAUSE:
1212 case KVM_TRACE_DISABLE:
1213 r = kvm_trace_ioctl(ioctl, arg);
1214 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001215 default:
Carsten Otte043405e2007-10-10 17:16:19 +02001216 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001217 }
1218out:
1219 return r;
1220}
1221
Avi Kivity6aa8b732006-12-10 02:21:36 -08001222static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08001223 .unlocked_ioctl = kvm_dev_ioctl,
1224 .compat_ioctl = kvm_dev_ioctl,
Avi Kivity6aa8b732006-12-10 02:21:36 -08001225};
1226
1227static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02001228 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08001229 "kvm",
1230 &kvm_chardev_ops,
1231};
1232
Avi Kivity1b6c0162007-05-24 13:03:52 +03001233static void hardware_enable(void *junk)
1234{
1235 int cpu = raw_smp_processor_id();
1236
1237 if (cpu_isset(cpu, cpus_hardware_enabled))
1238 return;
1239 cpu_set(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001240 kvm_arch_hardware_enable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03001241}
1242
1243static void hardware_disable(void *junk)
1244{
1245 int cpu = raw_smp_processor_id();
1246
1247 if (!cpu_isset(cpu, cpus_hardware_enabled))
1248 return;
1249 cpu_clear(cpu, cpus_hardware_enabled);
1250 decache_vcpus_on_cpu(cpu);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001251 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03001252}
1253
Avi Kivity774c47f2007-02-12 00:54:47 -08001254static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
1255 void *v)
1256{
1257 int cpu = (long)v;
1258
Avi Kivity1a6f4d72007-11-11 18:37:32 +02001259 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08001260 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03001261 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03001262 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
1263 cpu);
1264 hardware_disable(NULL);
1265 break;
Avi Kivity774c47f2007-02-12 00:54:47 -08001266 case CPU_UP_CANCELED:
Jeremy Katz43934a32007-02-19 14:37:46 +02001267 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
1268 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02001269 smp_call_function_single(cpu, hardware_disable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08001270 break;
Jeremy Katz43934a32007-02-19 14:37:46 +02001271 case CPU_ONLINE:
1272 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
1273 cpu);
Jens Axboe8691e5a2008-06-06 11:18:06 +02001274 smp_call_function_single(cpu, hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08001275 break;
1276 }
1277 return NOTIFY_OK;
1278}
1279
Rusty Russell9a2b85c2007-07-17 23:17:55 +10001280static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04001281 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10001282{
1283 if (val == SYS_RESTART) {
1284 /*
1285 * Some (well, at least mine) BIOSes hang on reboot if
1286 * in vmx root mode.
1287 */
1288 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
Jens Axboe15c8b6c2008-05-09 09:39:44 +02001289 on_each_cpu(hardware_disable, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10001290 }
1291 return NOTIFY_OK;
1292}
1293
1294static struct notifier_block kvm_reboot_notifier = {
1295 .notifier_call = kvm_reboot,
1296 .priority = 0,
1297};
1298
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04001299void kvm_io_bus_init(struct kvm_io_bus *bus)
1300{
1301 memset(bus, 0, sizeof(*bus));
1302}
1303
1304void kvm_io_bus_destroy(struct kvm_io_bus *bus)
1305{
1306 int i;
1307
1308 for (i = 0; i < bus->dev_count; i++) {
1309 struct kvm_io_device *pos = bus->devs[i];
1310
1311 kvm_iodevice_destructor(pos);
1312 }
1313}
1314
1315struct kvm_io_device *kvm_io_bus_find_dev(struct kvm_io_bus *bus, gpa_t addr)
1316{
1317 int i;
1318
1319 for (i = 0; i < bus->dev_count; i++) {
1320 struct kvm_io_device *pos = bus->devs[i];
1321
1322 if (pos->in_range(pos, addr))
1323 return pos;
1324 }
1325
1326 return NULL;
1327}
1328
1329void kvm_io_bus_register_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev)
1330{
1331 BUG_ON(bus->dev_count > (NR_IOBUS_DEVS-1));
1332
1333 bus->devs[bus->dev_count++] = dev;
1334}
1335
Avi Kivity774c47f2007-02-12 00:54:47 -08001336static struct notifier_block kvm_cpu_notifier = {
1337 .notifier_call = kvm_cpu_hotplug,
1338 .priority = 20, /* must be > scheduler priority */
1339};
1340
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001341static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02001342{
1343 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02001344 struct kvm *kvm;
1345
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001346 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02001347 spin_lock(&kvm_lock);
1348 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001349 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02001350 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001351 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02001352}
1353
1354DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
1355
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001356static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03001357{
1358 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03001359 struct kvm *kvm;
1360 struct kvm_vcpu *vcpu;
1361 int i;
1362
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001363 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03001364 spin_lock(&kvm_lock);
1365 list_for_each_entry(kvm, &vm_list, vm_list)
1366 for (i = 0; i < KVM_MAX_VCPUS; ++i) {
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001367 vcpu = kvm->vcpus[i];
1368 if (vcpu)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001369 *val += *(u32 *)((void *)vcpu + offset);
Avi Kivity1165f5f2007-04-19 17:27:43 +03001370 }
1371 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08001372 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03001373}
1374
Avi Kivityba1389b2007-11-18 16:24:12 +02001375DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
1376
1377static struct file_operations *stat_fops[] = {
1378 [KVM_STAT_VCPU] = &vcpu_stat_fops,
1379 [KVM_STAT_VM] = &vm_stat_fops,
1380};
Avi Kivity1165f5f2007-04-19 17:27:43 +03001381
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08001382static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001383{
1384 struct kvm_stats_debugfs_item *p;
1385
Hollis Blanchard76f7c872008-04-15 16:05:42 -05001386 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001387 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05001388 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03001389 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02001390 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001391}
1392
1393static void kvm_exit_debug(void)
1394{
1395 struct kvm_stats_debugfs_item *p;
1396
1397 for (p = debugfs_entries; p->name; ++p)
1398 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05001399 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001400}
1401
Avi Kivity59ae6c62007-02-12 00:54:48 -08001402static int kvm_suspend(struct sys_device *dev, pm_message_t state)
1403{
Avi Kivity4267c412007-05-24 13:09:41 +03001404 hardware_disable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08001405 return 0;
1406}
1407
1408static int kvm_resume(struct sys_device *dev)
1409{
Avi Kivity4267c412007-05-24 13:09:41 +03001410 hardware_enable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08001411 return 0;
1412}
1413
1414static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01001415 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08001416 .suspend = kvm_suspend,
1417 .resume = kvm_resume,
1418};
1419
1420static struct sys_device kvm_sysdev = {
1421 .id = 0,
1422 .cls = &kvm_sysdev_class,
1423};
1424
Izik Eiduscea7bb22007-10-17 19:17:48 +02001425struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05001426pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001427
Avi Kivity15ad7142007-07-11 18:17:21 +03001428static inline
1429struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
1430{
1431 return container_of(pn, struct kvm_vcpu, preempt_notifier);
1432}
1433
1434static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
1435{
1436 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
1437
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001438 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03001439}
1440
1441static void kvm_sched_out(struct preempt_notifier *pn,
1442 struct task_struct *next)
1443{
1444 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
1445
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001446 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03001447}
1448
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08001449int kvm_init(void *opaque, unsigned int vcpu_size,
Rusty Russellc16f8622007-07-30 21:12:19 +10001450 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001451{
1452 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03001453 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001454
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08001455 kvm_init_debug();
1456
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08001457 r = kvm_arch_init(opaque);
1458 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08001459 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08001460
1461 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
1462
1463 if (bad_page == NULL) {
1464 r = -ENOMEM;
1465 goto out;
1466 }
1467
Anthony Liguori35149e22008-04-02 14:46:56 -05001468 bad_pfn = page_to_pfn(bad_page);
1469
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001470 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001471 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08001472 goto out_free_0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001473
Yang, Sheng002c7f72007-07-31 14:23:01 +03001474 for_each_online_cpu(cpu) {
1475 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001476 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02001477 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03001478 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08001479 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03001480 }
1481
Jens Axboe15c8b6c2008-05-09 09:39:44 +02001482 on_each_cpu(hardware_enable, NULL, 1);
Avi Kivity774c47f2007-02-12 00:54:47 -08001483 r = register_cpu_notifier(&kvm_cpu_notifier);
1484 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08001485 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001486 register_reboot_notifier(&kvm_reboot_notifier);
1487
Avi Kivity59ae6c62007-02-12 00:54:48 -08001488 r = sysdev_class_register(&kvm_sysdev_class);
1489 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08001490 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08001491
1492 r = sysdev_register(&kvm_sysdev);
1493 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08001494 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08001495
Rusty Russellc16f8622007-07-30 21:12:19 +10001496 /* A kmem cache lets us meet the alignment requirements of fx_save. */
1497 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size,
Joe Perches56919c52007-11-12 20:06:51 -08001498 __alignof__(struct kvm_vcpu),
1499 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10001500 if (!kvm_vcpu_cache) {
1501 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08001502 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10001503 }
1504
Avi Kivity6aa8b732006-12-10 02:21:36 -08001505 kvm_chardev_ops.owner = module;
1506
1507 r = misc_register(&kvm_dev);
1508 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04001509 printk(KERN_ERR "kvm: misc device register failed\n");
Avi Kivity6aa8b732006-12-10 02:21:36 -08001510 goto out_free;
1511 }
1512
Avi Kivity15ad7142007-07-11 18:17:21 +03001513 kvm_preempt_ops.sched_in = kvm_sched_in;
1514 kvm_preempt_ops.sched_out = kvm_sched_out;
1515
Avi Kivityc7addb92007-09-16 18:58:32 +02001516 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001517
1518out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10001519 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08001520out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08001521 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08001522out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08001523 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08001524out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001525 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08001526 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08001527out_free_2:
Jens Axboe15c8b6c2008-05-09 09:39:44 +02001528 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaod23087842007-11-29 15:35:39 +08001529out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001530 kvm_arch_hardware_unsetup();
Zhang Xiantaod23087842007-11-29 15:35:39 +08001531out_free_0:
1532 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02001533out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08001534 kvm_arch_exit();
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08001535 kvm_exit_debug();
Zhang Xiantaod23087842007-11-29 15:35:39 +08001536out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08001537 return r;
1538}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08001539EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001540
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08001541void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001542{
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04001543 kvm_trace_cleanup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001544 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10001545 kmem_cache_destroy(kvm_vcpu_cache);
Avi Kivity59ae6c62007-02-12 00:54:48 -08001546 sysdev_unregister(&kvm_sysdev);
1547 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001548 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08001549 unregister_cpu_notifier(&kvm_cpu_notifier);
Jens Axboe15c8b6c2008-05-09 09:39:44 +02001550 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001551 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08001552 kvm_arch_exit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001553 kvm_exit_debug();
Izik Eiduscea7bb22007-10-17 19:17:48 +02001554 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001555}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08001556EXPORT_SYMBOL_GPL(kvm_exit);