blob: 5225052aebc15d28709359b7ee365f0490035203 [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.
Nicolas Kaiser9611c182010-10-06 14:23:22 +02008 * Copyright 2010 Red Hat, Inc. and/or its affiliates.
Avi Kivity6aa8b732006-12-10 02:21:36 -08009 *
10 * Authors:
11 * Avi Kivity <avi@qumranet.com>
12 * Yaniv Kamay <yaniv@qumranet.com>
13 *
14 * This work is licensed under the terms of the GNU GPL, version 2. See
15 * the COPYING file in the top-level directory.
16 *
17 */
18
Hollis Blancharde2174022007-12-03 15:30:24 -060019#include "iodev.h"
Avi Kivity6aa8b732006-12-10 02:21:36 -080020
Avi Kivityedf88412007-12-16 11:02:48 +020021#include <linux/kvm_host.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080022#include <linux/kvm.h>
23#include <linux/module.h>
24#include <linux/errno.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080025#include <linux/percpu.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>
Sheng Yange56d5322009-03-12 21:45:39 +080044#include <linux/bitops.h>
Marcelo Tosatti547de292009-05-07 17:55:13 -030045#include <linux/spinlock.h>
Arnd Bergmann6ff58942009-10-22 14:19:27 +020046#include <linux/compat.h>
Marcelo Tosattibc6678a2009-12-23 14:35:21 -020047#include <linux/srcu.h>
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +010048#include <linux/hugetlb.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090049#include <linux/slab.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080050
Avi Kivitye4956062007-06-28 14:15:57 -040051#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040052#include <asm/io.h>
53#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020054#include <asm/pgtable.h>
Alexander Grafc8240bd2009-10-30 05:47:26 +000055#include <asm-generic/bitops/le.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080056
Laurent Vivier5f94c172008-05-30 16:05:54 +020057#include "coalesced_mmio.h"
Laurent Vivier5f94c172008-05-30 16:05:54 +020058
Marcelo Tosatti229456f2009-06-17 09:22:14 -030059#define CREATE_TRACE_POINTS
60#include <trace/events/kvm.h>
61
Avi Kivity6aa8b732006-12-10 02:21:36 -080062MODULE_AUTHOR("Qumranet");
63MODULE_LICENSE("GPL");
64
Marcelo Tosattifa40a822009-06-04 15:08:24 -030065/*
66 * Ordering of locks:
67 *
Sheng Yangfae3a352009-12-15 10:28:07 +080068 * kvm->lock --> kvm->slots_lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030069 */
70
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080071DEFINE_SPINLOCK(kvm_lock);
72LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080073
Rusty Russell7f59f492008-12-07 21:25:45 +103074static cpumask_var_t cpus_hardware_enabled;
Alexander Graf10474ae2009-09-15 11:37:46 +020075static int kvm_usage_count = 0;
76static atomic_t hardware_enable_failed;
Avi Kivity1b6c0162007-05-24 13:03:52 +030077
Rusty Russellc16f8622007-07-30 21:12:19 +100078struct kmem_cache *kvm_vcpu_cache;
79EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030080
Avi Kivity15ad7142007-07-11 18:17:21 +030081static __read_mostly struct preempt_ops kvm_preempt_ops;
82
Hollis Blanchard76f7c872008-04-15 16:05:42 -050083struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080084
Avi Kivitybccf2152007-02-21 18:04:26 +020085static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
86 unsigned long arg);
Alexander Graf10474ae2009-09-15 11:37:46 +020087static int hardware_enable_all(void);
88static void hardware_disable_all(void);
Avi Kivitybccf2152007-02-21 18:04:26 +020089
Marcelo Tosattie93f8a02009-12-23 14:35:24 -020090static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
91
Hannes Edere8ba5d32008-11-28 17:02:42 +010092static bool kvm_rebooting;
Avi Kivity4ecac3f2008-05-13 13:23:38 +030093
Marcelo Tosatti54dee992009-06-11 12:07:44 -030094static bool largepages_enabled = true;
95
Gleb Natapovfa7bff82010-07-07 20:16:44 +030096static struct page *hwpoison_page;
97static pfn_t hwpoison_pfn;
Huang Yingbf998152010-05-31 14:28:19 +080098
Gleb Natapovedba23e2010-07-07 20:16:45 +030099static struct page *fault_page;
100static pfn_t fault_pfn;
101
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800102inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300103{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100104 if (pfn_valid(pfn)) {
105 struct page *page = compound_head(pfn_to_page(pfn));
106 return PageReserved(page);
107 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300108
109 return true;
110}
111
Avi Kivity6aa8b732006-12-10 02:21:36 -0800112/*
113 * Switches to specified vcpu, until a matching vcpu_put()
114 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200115void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800116{
Avi Kivity15ad7142007-07-11 18:17:21 +0300117 int cpu;
118
Avi Kivitybccf2152007-02-21 18:04:26 +0200119 mutex_lock(&vcpu->mutex);
Avi Kivity15ad7142007-07-11 18:17:21 +0300120 cpu = get_cpu();
121 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200122 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300123 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200124}
125
Carsten Otte313a3dc2007-10-11 19:16:52 +0200126void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800127{
Avi Kivity15ad7142007-07-11 18:17:21 +0300128 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200129 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300130 preempt_notifier_unregister(&vcpu->preempt_notifier);
131 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800132 mutex_unlock(&vcpu->mutex);
133}
134
Avi Kivityd9e368d2007-06-07 19:18:30 +0300135static void ack_flush(void *_completed)
136{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300137}
138
Rusty Russell49846892008-12-08 20:26:24 +1030139static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300140{
Avi Kivity597a5f52008-07-20 14:24:22 +0300141 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030142 cpumask_var_t cpus;
143 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300144 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300145
Li Zefan79f55992009-06-15 14:58:26 +0800146 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030147
Avi Kivity70e335e2010-02-18 11:25:22 +0200148 raw_spin_lock(&kvm->requests_lock);
Jan Kiszkae601e3b2009-07-20 11:30:12 +0200149 me = smp_processor_id();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300150 kvm_for_each_vcpu(i, vcpu, kvm) {
Avi Kivitya8eeb042010-05-10 12:34:53 +0300151 if (kvm_make_check_request(req, vcpu))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300152 continue;
153 cpu = vcpu->cpu;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030154 if (cpus != NULL && cpu != -1 && cpu != me)
155 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300156 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030157 if (unlikely(cpus == NULL))
158 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
159 else if (!cpumask_empty(cpus))
160 smp_call_function_many(cpus, ack_flush, NULL, 1);
161 else
162 called = false;
Avi Kivity70e335e2010-02-18 11:25:22 +0200163 raw_spin_unlock(&kvm->requests_lock);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030164 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030165 return called;
166}
167
168void kvm_flush_remote_tlbs(struct kvm *kvm)
169{
170 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
171 ++kvm->stat.remote_tlb_flush;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300172}
173
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500174void kvm_reload_remote_mmus(struct kvm *kvm)
175{
Rusty Russell49846892008-12-08 20:26:24 +1030176 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500177}
178
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000179int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
180{
181 struct page *page;
182 int r;
183
184 mutex_init(&vcpu->mutex);
185 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000186 vcpu->kvm = kvm;
187 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300188 init_waitqueue_head(&vcpu->wq);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000189
190 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
191 if (!page) {
192 r = -ENOMEM;
193 goto fail;
194 }
195 vcpu->run = page_address(page);
196
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800197 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000198 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800199 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000200 return 0;
201
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000202fail_free_run:
203 free_page((unsigned long)vcpu->run);
204fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000205 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000206}
207EXPORT_SYMBOL_GPL(kvm_vcpu_init);
208
209void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
210{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800211 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000212 free_page((unsigned long)vcpu->run);
213}
214EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
215
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200216#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
217static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
218{
219 return container_of(mn, struct kvm, mmu_notifier);
220}
221
222static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
223 struct mm_struct *mm,
224 unsigned long address)
225{
226 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200227 int need_tlb_flush, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200228
229 /*
230 * When ->invalidate_page runs, the linux pte has been zapped
231 * already but the page is still allocated until
232 * ->invalidate_page returns. So if we increase the sequence
233 * here the kvm page fault will notice if the spte can't be
234 * established because the page is going to be freed. If
235 * instead the kvm page fault establishes the spte before
236 * ->invalidate_page runs, kvm_unmap_hva will release it
237 * before returning.
238 *
239 * The sequence increase only need to be seen at spin_unlock
240 * time, and not at spin_lock time.
241 *
242 * Increasing the sequence after the spin_unlock would be
243 * unsafe because the kvm page fault could then establish the
244 * pte after kvm_unmap_hva returned, without noticing the page
245 * is going to be freed.
246 */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200247 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200248 spin_lock(&kvm->mmu_lock);
249 kvm->mmu_notifier_seq++;
250 need_tlb_flush = kvm_unmap_hva(kvm, address);
251 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200252 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200253
254 /* we've to flush the tlb before the pages can be freed */
255 if (need_tlb_flush)
256 kvm_flush_remote_tlbs(kvm);
257
258}
259
Izik Eidus3da0dd42009-09-23 21:47:18 +0300260static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
261 struct mm_struct *mm,
262 unsigned long address,
263 pte_t pte)
264{
265 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200266 int idx;
Izik Eidus3da0dd42009-09-23 21:47:18 +0300267
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200268 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300269 spin_lock(&kvm->mmu_lock);
270 kvm->mmu_notifier_seq++;
271 kvm_set_spte_hva(kvm, address, pte);
272 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200273 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300274}
275
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200276static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
277 struct mm_struct *mm,
278 unsigned long start,
279 unsigned long end)
280{
281 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200282 int need_tlb_flush = 0, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200283
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200284 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200285 spin_lock(&kvm->mmu_lock);
286 /*
287 * The count increase must become visible at unlock time as no
288 * spte can be established without taking the mmu_lock and
289 * count is also read inside the mmu_lock critical section.
290 */
291 kvm->mmu_notifier_count++;
292 for (; start < end; start += PAGE_SIZE)
293 need_tlb_flush |= kvm_unmap_hva(kvm, start);
294 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200295 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200296
297 /* we've to flush the tlb before the pages can be freed */
298 if (need_tlb_flush)
299 kvm_flush_remote_tlbs(kvm);
300}
301
302static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
303 struct mm_struct *mm,
304 unsigned long start,
305 unsigned long end)
306{
307 struct kvm *kvm = mmu_notifier_to_kvm(mn);
308
309 spin_lock(&kvm->mmu_lock);
310 /*
311 * This sequence increase will notify the kvm page fault that
312 * the page that is going to be mapped in the spte could have
313 * been freed.
314 */
315 kvm->mmu_notifier_seq++;
316 /*
317 * The above sequence increase must be visible before the
318 * below count decrease but both values are read by the kvm
319 * page fault under mmu_lock spinlock so we don't need to add
320 * a smb_wmb() here in between the two.
321 */
322 kvm->mmu_notifier_count--;
323 spin_unlock(&kvm->mmu_lock);
324
325 BUG_ON(kvm->mmu_notifier_count < 0);
326}
327
328static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
329 struct mm_struct *mm,
330 unsigned long address)
331{
332 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200333 int young, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200334
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200335 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200336 spin_lock(&kvm->mmu_lock);
337 young = kvm_age_hva(kvm, address);
338 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200339 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200340
341 if (young)
342 kvm_flush_remote_tlbs(kvm);
343
344 return young;
345}
346
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100347static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
348 struct mm_struct *mm)
349{
350 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800351 int idx;
352
353 idx = srcu_read_lock(&kvm->srcu);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100354 kvm_arch_flush_shadow(kvm);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800355 srcu_read_unlock(&kvm->srcu, idx);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100356}
357
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200358static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
359 .invalidate_page = kvm_mmu_notifier_invalidate_page,
360 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
361 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
362 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300363 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100364 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200365};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200366
367static int kvm_init_mmu_notifier(struct kvm *kvm)
368{
369 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
370 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
371}
372
373#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
374
375static int kvm_init_mmu_notifier(struct kvm *kvm)
376{
377 return 0;
378}
379
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200380#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
381
Avi Kivityf17abe92007-02-21 19:28:04 +0200382static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800383{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200384 int r = 0, i;
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800385 struct kvm *kvm = kvm_arch_create_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800386
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800387 if (IS_ERR(kvm))
388 goto out;
Alexander Graf10474ae2009-09-15 11:37:46 +0200389
390 r = hardware_enable_all();
391 if (r)
392 goto out_err_nodisable;
393
Avi Kivity75858a82009-01-04 17:10:50 +0200394#ifdef CONFIG_HAVE_KVM_IRQCHIP
395 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300396 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200397#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800398
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200399 r = -ENOMEM;
400 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
401 if (!kvm->memslots)
402 goto out_err;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200403 if (init_srcu_struct(&kvm->srcu))
404 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200405 for (i = 0; i < KVM_NR_BUSES; i++) {
406 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
407 GFP_KERNEL);
408 if (!kvm->buses[i]) {
409 cleanup_srcu_struct(&kvm->srcu);
410 goto out_err;
411 }
412 }
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200413
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200414 r = kvm_init_mmu_notifier(kvm);
Avi Kivity283d0c62009-12-20 14:25:19 +0200415 if (r) {
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200416 cleanup_srcu_struct(&kvm->srcu);
Avi Kivity283d0c62009-12-20 14:25:19 +0200417 goto out_err;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200418 }
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200419
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200420 kvm->mm = current->mm;
421 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500422 spin_lock_init(&kvm->mmu_lock);
Avi Kivity70e335e2010-02-18 11:25:22 +0200423 raw_spin_lock_init(&kvm->requests_lock);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400424 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800425 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300426 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200427 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300428 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000429 spin_lock(&kvm_lock);
430 list_add(&kvm->vm_list, &vm_list);
431 spin_unlock(&kvm_lock);
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800432out:
Avi Kivityf17abe92007-02-21 19:28:04 +0200433 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200434
435out_err:
436 hardware_disable_all();
437out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200438 for (i = 0; i < KVM_NR_BUSES; i++)
439 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200440 kfree(kvm->memslots);
Alexander Graf10474ae2009-09-15 11:37:46 +0200441 kfree(kvm);
442 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200443}
444
Avi Kivity6aa8b732006-12-10 02:21:36 -0800445/*
446 * Free any memory in @free but not in @dont.
447 */
448static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
449 struct kvm_memory_slot *dont)
450{
Joerg Roedelec04b262009-06-19 15:16:23 +0200451 int i;
452
Izik Eidus290fc382007-09-27 14:11:22 +0200453 if (!dont || free->rmap != dont->rmap)
454 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800455
456 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
457 vfree(free->dirty_bitmap);
458
Joerg Roedelec04b262009-06-19 15:16:23 +0200459
460 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
461 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
462 vfree(free->lpage_info[i]);
463 free->lpage_info[i] = NULL;
464 }
465 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300466
Avi Kivity6aa8b732006-12-10 02:21:36 -0800467 free->npages = 0;
Al Viro8b6d44c2007-02-09 16:38:40 +0000468 free->dirty_bitmap = NULL;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500469 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800470}
471
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800472void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800473{
474 int i;
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200475 struct kvm_memslots *slots = kvm->memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800476
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200477 for (i = 0; i < slots->nmemslots; ++i)
478 kvm_free_physmem_slot(&slots->memslots[i], NULL);
479
480 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800481}
482
Avi Kivityf17abe92007-02-21 19:28:04 +0200483static void kvm_destroy_vm(struct kvm *kvm)
484{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200485 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200486 struct mm_struct *mm = kvm->mm;
487
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800488 kvm_arch_sync_events(kvm);
Avi Kivity133de902007-02-12 00:54:44 -0800489 spin_lock(&kvm_lock);
490 list_del(&kvm->vm_list);
491 spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200492 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200493 for (i = 0; i < KVM_NR_BUSES; i++)
494 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200495 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200496#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
497 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200498#else
499 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200500#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800501 kvm_arch_destroy_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200502 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200503 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200504}
505
Izik Eidusd39f13b2008-03-30 16:01:25 +0300506void kvm_get_kvm(struct kvm *kvm)
507{
508 atomic_inc(&kvm->users_count);
509}
510EXPORT_SYMBOL_GPL(kvm_get_kvm);
511
512void kvm_put_kvm(struct kvm *kvm)
513{
514 if (atomic_dec_and_test(&kvm->users_count))
515 kvm_destroy_vm(kvm);
516}
517EXPORT_SYMBOL_GPL(kvm_put_kvm);
518
519
Avi Kivityf17abe92007-02-21 19:28:04 +0200520static int kvm_vm_release(struct inode *inode, struct file *filp)
521{
522 struct kvm *kvm = filp->private_data;
523
Gregory Haskins721eecb2009-05-20 10:30:49 -0400524 kvm_irqfd_release(kvm);
525
Izik Eidusd39f13b2008-03-30 16:01:25 +0300526 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800527 return 0;
528}
529
Avi Kivity6aa8b732006-12-10 02:21:36 -0800530/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800531 * Allocate some memory and give it an address in the guest physical address
532 * space.
533 *
534 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800535 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500536 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800537 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800538int __kvm_set_memory_region(struct kvm *kvm,
539 struct kvm_userspace_memory_region *mem,
540 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800541{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200542 int r, flush_shadow = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800543 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200544 unsigned long npages;
545 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800546 struct kvm_memory_slot *memslot;
547 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200548 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800549
550 r = -EINVAL;
551 /* General sanity checks */
552 if (mem->memory_size & (PAGE_SIZE - 1))
553 goto out;
554 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
555 goto out;
Sheng Yange7cacd42008-11-11 15:30:40 +0800556 if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600557 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200558 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800559 goto out;
560 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
561 goto out;
562
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200563 memslot = &kvm->memslots->memslots[mem->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800564 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
565 npages = mem->memory_size >> PAGE_SHIFT;
566
Takuya Yoshikawa660c22c2010-04-13 22:47:24 +0900567 r = -EINVAL;
568 if (npages > KVM_MEM_MAX_NR_PAGES)
569 goto out;
570
Avi Kivity6aa8b732006-12-10 02:21:36 -0800571 if (!npages)
572 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
573
Avi Kivity6aa8b732006-12-10 02:21:36 -0800574 new = old = *memslot;
575
Avi Kivitye36d96f2010-06-21 10:56:36 +0300576 new.id = mem->slot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800577 new.base_gfn = base_gfn;
578 new.npages = npages;
579 new.flags = mem->flags;
580
581 /* Disallow changing a memory slot's size. */
582 r = -EINVAL;
583 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800584 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800585
586 /* Check for overlaps */
587 r = -EEXIST;
588 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200589 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800590
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200591 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800592 continue;
593 if (!((base_gfn + npages <= s->base_gfn) ||
594 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800595 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800596 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800597
Avi Kivity6aa8b732006-12-10 02:21:36 -0800598 /* Free page dirty bitmap if unneeded */
599 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000600 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800601
602 r = -ENOMEM;
603
604 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +0200605#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500606 if (npages && !new.rmap) {
Lai Jiangshan3bd89002010-06-02 17:01:23 +0800607 new.rmap = vmalloc(npages * sizeof(*new.rmap));
Izik Eidus290fc382007-09-27 14:11:22 +0200608
609 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800610 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200611
Izik Eidus290fc382007-09-27 14:11:22 +0200612 memset(new.rmap, 0, npages * sizeof(*new.rmap));
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500613
Izik Eidus80b14b52007-10-25 11:54:04 +0200614 new.user_alloc = user_alloc;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200615 new.userspace_addr = mem->userspace_addr;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800616 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200617 if (!npages)
618 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300619
Joerg Roedelec04b262009-06-19 15:16:23 +0200620 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +0200621 unsigned long ugfn;
622 unsigned long j;
623 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +0200624 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300625
Joerg Roedelec04b262009-06-19 15:16:23 +0200626 /* Avoid unused variable warning if no large pages */
627 (void)level;
628
629 if (new.lpage_info[i])
630 continue;
631
Joerg Roedel82855412010-07-01 16:00:11 +0200632 lpages = 1 + ((base_gfn + npages - 1)
633 >> KVM_HPAGE_GFN_SHIFT(level));
634 lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level);
Joerg Roedelec04b262009-06-19 15:16:23 +0200635
636 new.lpage_info[i] = vmalloc(lpages * sizeof(*new.lpage_info[i]));
637
638 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300639 goto out_free;
640
Joerg Roedelec04b262009-06-19 15:16:23 +0200641 memset(new.lpage_info[i], 0,
642 lpages * sizeof(*new.lpage_info[i]));
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300643
Joerg Roedel82855412010-07-01 16:00:11 +0200644 if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200645 new.lpage_info[i][0].write_count = 1;
Joerg Roedel82855412010-07-01 16:00:11 +0200646 if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200647 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +0300648 ugfn = new.userspace_addr >> PAGE_SHIFT;
649 /*
650 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300651 * other, or if explicitly asked to, disable large page
652 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +0300653 */
Joerg Roedelec04b262009-06-19 15:16:23 +0200654 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300655 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +0200656 for (j = 0; j < lpages; ++j)
657 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300658 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800659
Joerg Roedelec04b262009-06-19 15:16:23 +0200660skip_lpage:
661
Avi Kivity6aa8b732006-12-10 02:21:36 -0800662 /* Allocate page dirty bitmap if needed */
663 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900664 unsigned long dirty_bytes = kvm_dirty_bitmap_bytes(&new);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800665
666 new.dirty_bitmap = vmalloc(dirty_bytes);
667 if (!new.dirty_bitmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800668 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800669 memset(new.dirty_bitmap, 0, dirty_bytes);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200670 /* destroy any largepage mappings for dirty tracking */
Izik Eiduse2445842009-06-10 19:23:24 +0300671 if (old.npages)
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200672 flush_shadow = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800673 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +0200674#else /* not defined CONFIG_S390 */
675 new.user_alloc = user_alloc;
676 if (user_alloc)
677 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200678#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800679
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200680 if (!npages) {
681 r = -ENOMEM;
682 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
683 if (!slots)
684 goto out_free;
685 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
686 if (mem->slot >= slots->nmemslots)
687 slots->nmemslots = mem->slot + 1;
688 slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID;
689
690 old_memslots = kvm->memslots;
691 rcu_assign_pointer(kvm->memslots, slots);
692 synchronize_srcu_expedited(&kvm->srcu);
693 /* From this point no new shadow pages pointing to a deleted
694 * memslot will be created.
695 *
696 * validation of sp->gfn happens in:
697 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
698 * - kvm_is_visible_gfn (mmu_check_roots)
699 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300700 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200701 kfree(old_memslots);
702 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300703
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200704 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
705 if (r)
706 goto out_free;
707
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200708 /* map the pages in iommu page table */
709 if (npages) {
710 r = kvm_iommu_map_pages(kvm, &new);
711 if (r)
712 goto out_free;
713 }
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200714
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200715 r = -ENOMEM;
716 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
717 if (!slots)
718 goto out_free;
719 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
720 if (mem->slot >= slots->nmemslots)
721 slots->nmemslots = mem->slot + 1;
722
723 /* actual memory is freed via old in kvm_free_physmem_slot below */
724 if (!npages) {
725 new.rmap = NULL;
726 new.dirty_bitmap = NULL;
727 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
728 new.lpage_info[i] = NULL;
729 }
730
731 slots->memslots[mem->slot] = new;
732 old_memslots = kvm->memslots;
733 rcu_assign_pointer(kvm->memslots, slots);
734 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800735
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200736 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800737
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200738 kvm_free_physmem_slot(&old, &new);
739 kfree(old_memslots);
740
741 if (flush_shadow)
742 kvm_arch_flush_shadow(kvm);
743
Avi Kivity6aa8b732006-12-10 02:21:36 -0800744 return 0;
745
Sheng Yangf78e0e22007-10-29 09:40:42 +0800746out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800747 kvm_free_physmem_slot(&new, &old);
748out:
749 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200750
751}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800752EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
753
754int kvm_set_memory_region(struct kvm *kvm,
755 struct kvm_userspace_memory_region *mem,
756 int user_alloc)
757{
758 int r;
759
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200760 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800761 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200762 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800763 return r;
764}
Izik Eidus210c7c42007-10-24 23:52:57 +0200765EXPORT_SYMBOL_GPL(kvm_set_memory_region);
766
Carsten Otte1fe779f2007-10-29 16:08:35 +0100767int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
768 struct
769 kvm_userspace_memory_region *mem,
770 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200771{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200772 if (mem->slot >= KVM_MEMORY_SLOTS)
773 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200774 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800775}
776
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800777int kvm_get_dirty_log(struct kvm *kvm,
778 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800779{
780 struct kvm_memory_slot *memslot;
781 int r, i;
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900782 unsigned long n;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800783 unsigned long any = 0;
784
Avi Kivity6aa8b732006-12-10 02:21:36 -0800785 r = -EINVAL;
786 if (log->slot >= KVM_MEMORY_SLOTS)
787 goto out;
788
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200789 memslot = &kvm->memslots->memslots[log->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800790 r = -ENOENT;
791 if (!memslot->dirty_bitmap)
792 goto out;
793
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900794 n = kvm_dirty_bitmap_bytes(memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800795
Uri Lublincd1a4a92007-02-22 16:43:09 +0200796 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800797 any = memslot->dirty_bitmap[i];
798
799 r = -EFAULT;
800 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
801 goto out;
802
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800803 if (any)
804 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800805
806 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800807out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800808 return r;
809}
810
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300811void kvm_disable_largepages(void)
812{
813 largepages_enabled = false;
814}
815EXPORT_SYMBOL_GPL(kvm_disable_largepages);
816
Izik Eiduscea7bb22007-10-17 19:17:48 +0200817int is_error_page(struct page *page)
818{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300819 return page == bad_page || page == hwpoison_page || page == fault_page;
Izik Eiduscea7bb22007-10-17 19:17:48 +0200820}
821EXPORT_SYMBOL_GPL(is_error_page);
822
Anthony Liguori35149e22008-04-02 14:46:56 -0500823int is_error_pfn(pfn_t pfn)
824{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300825 return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500826}
827EXPORT_SYMBOL_GPL(is_error_pfn);
828
Huang Yingbf998152010-05-31 14:28:19 +0800829int is_hwpoison_pfn(pfn_t pfn)
830{
831 return pfn == hwpoison_pfn;
832}
833EXPORT_SYMBOL_GPL(is_hwpoison_pfn);
834
Gleb Natapovedba23e2010-07-07 20:16:45 +0300835int is_fault_pfn(pfn_t pfn)
836{
837 return pfn == fault_pfn;
838}
839EXPORT_SYMBOL_GPL(is_fault_pfn);
840
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200841static inline unsigned long bad_hva(void)
842{
843 return PAGE_OFFSET;
844}
845
846int kvm_is_error_hva(unsigned long addr)
847{
848 return addr == bad_hva();
849}
850EXPORT_SYMBOL_GPL(kvm_is_error_hva);
851
Avi Kivitya1f4d392010-06-21 11:44:20 +0300852struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800853{
854 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +0800855 struct kvm_memslots *slots = kvm_memslots(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800856
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200857 for (i = 0; i < slots->nmemslots; ++i) {
858 struct kvm_memory_slot *memslot = &slots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800859
860 if (gfn >= memslot->base_gfn
861 && gfn < memslot->base_gfn + memslot->npages)
862 return memslot;
863 }
Al Viro8b6d44c2007-02-09 16:38:40 +0000864 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800865}
Avi Kivitya1f4d392010-06-21 11:44:20 +0300866EXPORT_SYMBOL_GPL(gfn_to_memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800867
Izik Eiduse0d62c72007-10-24 23:57:46 +0200868int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
869{
870 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +0800871 struct kvm_memslots *slots = kvm_memslots(kvm);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200872
Izik Eiduse0d62c72007-10-24 23:57:46 +0200873 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200874 struct kvm_memory_slot *memslot = &slots->memslots[i];
Izik Eiduse0d62c72007-10-24 23:57:46 +0200875
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200876 if (memslot->flags & KVM_MEMSLOT_INVALID)
877 continue;
878
Izik Eiduse0d62c72007-10-24 23:57:46 +0200879 if (gfn >= memslot->base_gfn
880 && gfn < memslot->base_gfn + memslot->npages)
881 return 1;
882 }
883 return 0;
884}
885EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
886
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +0100887unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
888{
889 struct vm_area_struct *vma;
890 unsigned long addr, size;
891
892 size = PAGE_SIZE;
893
894 addr = gfn_to_hva(kvm, gfn);
895 if (kvm_is_error_hva(addr))
896 return PAGE_SIZE;
897
898 down_read(&current->mm->mmap_sem);
899 vma = find_vma(current->mm, addr);
900 if (!vma)
901 goto out;
902
903 size = vma_kernel_pagesize(vma);
904
905out:
906 up_read(&current->mm->mmap_sem);
907
908 return size;
909}
910
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200911int memslot_id(struct kvm *kvm, gfn_t gfn)
912{
913 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +0800914 struct kvm_memslots *slots = kvm_memslots(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200915 struct kvm_memory_slot *memslot = NULL;
916
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200917 for (i = 0; i < slots->nmemslots; ++i) {
918 memslot = &slots->memslots[i];
919
920 if (gfn >= memslot->base_gfn
921 && gfn < memslot->base_gfn + memslot->npages)
922 break;
923 }
924
925 return memslot - slots->memslots;
926}
927
Xiao Guangrong48987782010-08-22 19:11:43 +0800928static unsigned long gfn_to_hva_many(struct kvm *kvm, gfn_t gfn,
929 gfn_t *nr_pages)
Izik Eidus539cb662007-11-11 22:05:04 +0200930{
931 struct kvm_memory_slot *slot;
932
Avi Kivitya1f4d392010-06-21 11:44:20 +0300933 slot = gfn_to_memslot(kvm, gfn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200934 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +0200935 return bad_hva();
Xiao Guangrong48987782010-08-22 19:11:43 +0800936
937 if (nr_pages)
938 *nr_pages = slot->npages - (gfn - slot->base_gfn);
939
Takuya Yoshikawaf5c98032010-02-25 11:33:19 +0900940 return gfn_to_hva_memslot(slot, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +0200941}
Xiao Guangrong48987782010-08-22 19:11:43 +0800942
943unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
944{
945 return gfn_to_hva_many(kvm, gfn, NULL);
946}
Sheng Yang0d150292008-04-25 21:44:50 +0800947EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +0200948
Xiao Guangrong887c08a2010-08-22 19:10:28 +0800949static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic)
Avi Kivity954bbbc22007-03-30 14:02:32 +0300950{
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500951 struct page *page[1];
952 int npages;
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500953 pfn_t pfn;
Avi Kivity954bbbc22007-03-30 14:02:32 +0300954
Xiao Guangrong887c08a2010-08-22 19:10:28 +0800955 if (atomic)
956 npages = __get_user_pages_fast(addr, 1, 1, page);
957 else {
958 might_sleep();
959 npages = get_user_pages_fast(addr, 1, 1, page);
960 }
Izik Eidus539cb662007-11-11 22:05:04 +0200961
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500962 if (unlikely(npages != 1)) {
963 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500964
Xiao Guangrong887c08a2010-08-22 19:10:28 +0800965 if (atomic)
966 goto return_fault_page;
967
Huang Yingbbeb3402010-06-22 14:23:11 +0800968 down_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +0800969 if (is_hwpoison_address(addr)) {
Huang Yingbbeb3402010-06-22 14:23:11 +0800970 up_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +0800971 get_page(hwpoison_page);
972 return page_to_pfn(hwpoison_page);
973 }
974
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500975 vma = find_vma(current->mm, addr);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300976
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500977 if (vma == NULL || addr < vma->vm_start ||
978 !(vma->vm_flags & VM_PFNMAP)) {
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300979 up_read(&current->mm->mmap_sem);
Xiao Guangrong887c08a2010-08-22 19:10:28 +0800980return_fault_page:
Gleb Natapovedba23e2010-07-07 20:16:45 +0300981 get_page(fault_page);
982 return page_to_pfn(fault_page);
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500983 }
984
985 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -0300986 up_read(&current->mm->mmap_sem);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800987 BUG_ON(!kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -0500988 } else
989 pfn = page_to_pfn(page[0]);
990
991 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500992}
993
Xiao Guangrong887c08a2010-08-22 19:10:28 +0800994pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr)
995{
996 return hva_to_pfn(kvm, addr, true);
997}
998EXPORT_SYMBOL_GPL(hva_to_pfn_atomic);
999
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001000static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic)
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001001{
1002 unsigned long addr;
1003
1004 addr = gfn_to_hva(kvm, gfn);
1005 if (kvm_is_error_hva(addr)) {
1006 get_page(bad_page);
1007 return page_to_pfn(bad_page);
1008 }
1009
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001010 return hva_to_pfn(kvm, addr, atomic);
1011}
1012
1013pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
1014{
1015 return __gfn_to_pfn(kvm, gfn, true);
1016}
1017EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);
1018
1019pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
1020{
1021 return __gfn_to_pfn(kvm, gfn, false);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001022}
Anthony Liguori35149e22008-04-02 14:46:56 -05001023EXPORT_SYMBOL_GPL(gfn_to_pfn);
1024
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001025pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
1026 struct kvm_memory_slot *slot, gfn_t gfn)
1027{
1028 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001029 return hva_to_pfn(kvm, addr, false);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001030}
1031
Xiao Guangrong48987782010-08-22 19:11:43 +08001032int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
1033 int nr_pages)
1034{
1035 unsigned long addr;
1036 gfn_t entry;
1037
1038 addr = gfn_to_hva_many(kvm, gfn, &entry);
1039 if (kvm_is_error_hva(addr))
1040 return -1;
1041
1042 if (entry < nr_pages)
1043 return 0;
1044
1045 return __get_user_pages_fast(addr, nr_pages, 1, pages);
1046}
1047EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);
1048
Anthony Liguori35149e22008-04-02 14:46:56 -05001049struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1050{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001051 pfn_t pfn;
1052
1053 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001054 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001055 return pfn_to_page(pfn);
1056
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001057 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001058
1059 get_page(bad_page);
1060 return bad_page;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001061}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001062
Avi Kivity954bbbc22007-03-30 14:02:32 +03001063EXPORT_SYMBOL_GPL(gfn_to_page);
1064
Izik Eidusb4231d62007-11-20 11:49:33 +02001065void kvm_release_page_clean(struct page *page)
1066{
Anthony Liguori35149e22008-04-02 14:46:56 -05001067 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001068}
1069EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1070
Anthony Liguori35149e22008-04-02 14:46:56 -05001071void kvm_release_pfn_clean(pfn_t pfn)
1072{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001073 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001074 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001075}
1076EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1077
Izik Eidusb4231d62007-11-20 11:49:33 +02001078void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001079{
Anthony Liguori35149e22008-04-02 14:46:56 -05001080 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001081}
Izik Eidusb4231d62007-11-20 11:49:33 +02001082EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001083
Anthony Liguori35149e22008-04-02 14:46:56 -05001084void kvm_release_pfn_dirty(pfn_t pfn)
1085{
1086 kvm_set_pfn_dirty(pfn);
1087 kvm_release_pfn_clean(pfn);
1088}
1089EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1090
1091void kvm_set_page_dirty(struct page *page)
1092{
1093 kvm_set_pfn_dirty(page_to_pfn(page));
1094}
1095EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1096
1097void kvm_set_pfn_dirty(pfn_t pfn)
1098{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001099 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001100 struct page *page = pfn_to_page(pfn);
1101 if (!PageReserved(page))
1102 SetPageDirty(page);
1103 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001104}
1105EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1106
1107void kvm_set_pfn_accessed(pfn_t pfn)
1108{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001109 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001110 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001111}
1112EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1113
1114void kvm_get_pfn(pfn_t pfn)
1115{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001116 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001117 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001118}
1119EXPORT_SYMBOL_GPL(kvm_get_pfn);
1120
Izik Eidus195aefd2007-10-01 22:14:18 +02001121static int next_segment(unsigned long len, int offset)
1122{
1123 if (len > PAGE_SIZE - offset)
1124 return PAGE_SIZE - offset;
1125 else
1126 return len;
1127}
1128
1129int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1130 int len)
1131{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001132 int r;
1133 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001134
Izik Eiduse0506bc2007-11-11 22:10:22 +02001135 addr = gfn_to_hva(kvm, gfn);
1136 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001137 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001138 r = copy_from_user(data, (void __user *)addr + offset, len);
1139 if (r)
1140 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001141 return 0;
1142}
1143EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1144
1145int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1146{
1147 gfn_t gfn = gpa >> PAGE_SHIFT;
1148 int seg;
1149 int offset = offset_in_page(gpa);
1150 int ret;
1151
1152 while ((seg = next_segment(len, offset)) != 0) {
1153 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1154 if (ret < 0)
1155 return ret;
1156 offset = 0;
1157 len -= seg;
1158 data += seg;
1159 ++gfn;
1160 }
1161 return 0;
1162}
1163EXPORT_SYMBOL_GPL(kvm_read_guest);
1164
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001165int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1166 unsigned long len)
1167{
1168 int r;
1169 unsigned long addr;
1170 gfn_t gfn = gpa >> PAGE_SHIFT;
1171 int offset = offset_in_page(gpa);
1172
1173 addr = gfn_to_hva(kvm, gfn);
1174 if (kvm_is_error_hva(addr))
1175 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001176 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001177 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001178 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001179 if (r)
1180 return -EFAULT;
1181 return 0;
1182}
1183EXPORT_SYMBOL(kvm_read_guest_atomic);
1184
Izik Eidus195aefd2007-10-01 22:14:18 +02001185int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1186 int offset, int len)
1187{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001188 int r;
1189 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001190
Izik Eiduse0506bc2007-11-11 22:10:22 +02001191 addr = gfn_to_hva(kvm, gfn);
1192 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001193 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001194 r = copy_to_user((void __user *)addr + offset, data, len);
1195 if (r)
1196 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001197 mark_page_dirty(kvm, gfn);
1198 return 0;
1199}
1200EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1201
1202int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1203 unsigned long len)
1204{
1205 gfn_t gfn = gpa >> PAGE_SHIFT;
1206 int seg;
1207 int offset = offset_in_page(gpa);
1208 int ret;
1209
1210 while ((seg = next_segment(len, offset)) != 0) {
1211 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1212 if (ret < 0)
1213 return ret;
1214 offset = 0;
1215 len -= seg;
1216 data += seg;
1217 ++gfn;
1218 }
1219 return 0;
1220}
1221
1222int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1223{
Izik Eidus3e021bf2007-11-19 11:16:57 +02001224 return kvm_write_guest_page(kvm, gfn, empty_zero_page, offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001225}
1226EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1227
1228int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1229{
1230 gfn_t gfn = gpa >> PAGE_SHIFT;
1231 int seg;
1232 int offset = offset_in_page(gpa);
1233 int ret;
1234
1235 while ((seg = next_segment(len, offset)) != 0) {
1236 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1237 if (ret < 0)
1238 return ret;
1239 offset = 0;
1240 len -= seg;
1241 ++gfn;
1242 }
1243 return 0;
1244}
1245EXPORT_SYMBOL_GPL(kvm_clear_guest);
1246
Avi Kivity6aa8b732006-12-10 02:21:36 -08001247void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1248{
Nguyen Anh Quynh31389942007-06-05 10:35:19 +03001249 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001250
Avi Kivitya1f4d392010-06-21 11:44:20 +03001251 memslot = gfn_to_memslot(kvm, gfn);
Rusty Russell7e9d6192007-07-31 20:41:14 +10001252 if (memslot && memslot->dirty_bitmap) {
1253 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001254
Takuya Yoshikawad1476932010-04-23 17:48:35 +09001255 generic___set_le_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001256 }
1257}
1258
Eddie Dongb6958ce2007-07-18 12:15:21 +03001259/*
1260 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1261 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001262void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001263{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001264 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001265
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001266 for (;;) {
1267 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001268
Gleb Natapova1b37102009-07-09 15:33:52 +03001269 if (kvm_arch_vcpu_runnable(vcpu)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001270 kvm_make_request(KVM_REQ_UNHALT, vcpu);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001271 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001272 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001273 if (kvm_cpu_has_pending_timer(vcpu))
1274 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001275 if (signal_pending(current))
1276 break;
1277
Eddie Dongb6958ce2007-07-18 12:15:21 +03001278 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001279 }
1280
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001281 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001282}
1283
Avi Kivity6aa8b732006-12-10 02:21:36 -08001284void kvm_resched(struct kvm_vcpu *vcpu)
1285{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001286 if (!need_resched())
1287 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001288 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001289}
1290EXPORT_SYMBOL_GPL(kvm_resched);
1291
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001292void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu)
1293{
1294 ktime_t expires;
1295 DEFINE_WAIT(wait);
1296
1297 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
1298
1299 /* Sleep for 100 us, and hope lock-holder got scheduled */
1300 expires = ktime_add_ns(ktime_get(), 100000UL);
1301 schedule_hrtimeout(&expires, HRTIMER_MODE_ABS);
1302
1303 finish_wait(&vcpu->wq, &wait);
1304}
1305EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1306
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001307static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001308{
1309 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001310 struct page *page;
1311
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001312 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001313 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001314#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001315 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001316 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001317#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001318#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1319 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1320 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1321#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001322 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001323 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001324 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001325 vmf->page = page;
1326 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001327}
1328
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001329static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001330 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001331};
1332
1333static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1334{
1335 vma->vm_ops = &kvm_vcpu_vm_ops;
1336 return 0;
1337}
1338
Avi Kivitybccf2152007-02-21 18:04:26 +02001339static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1340{
1341 struct kvm_vcpu *vcpu = filp->private_data;
1342
Al Viro66c0b392008-04-19 20:33:56 +01001343 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001344 return 0;
1345}
1346
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001347static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001348 .release = kvm_vcpu_release,
1349 .unlocked_ioctl = kvm_vcpu_ioctl,
1350 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001351 .mmap = kvm_vcpu_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001352 .llseek = noop_llseek,
Avi Kivitybccf2152007-02-21 18:04:26 +02001353};
1354
1355/*
1356 * Allocates an inode for the vcpu.
1357 */
1358static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1359{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001360 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001361}
1362
Avi Kivityc5ea7662007-02-20 18:41:05 +02001363/*
1364 * Creates some virtual cpus. Good luck creating more than one.
1365 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001366static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001367{
1368 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001369 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001370
Gleb Natapov73880c82009-06-09 15:56:28 +03001371 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001372 if (IS_ERR(vcpu))
1373 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001374
Avi Kivity15ad7142007-07-11 18:17:21 +03001375 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1376
Avi Kivity26e52152007-11-20 15:30:24 +02001377 r = kvm_arch_vcpu_setup(vcpu);
1378 if (r)
Glauber Costa7d8fece2008-09-17 23:16:59 -03001379 return r;
Avi Kivity26e52152007-11-20 15:30:24 +02001380
Shaohua Li11ec2802007-07-23 14:51:37 +08001381 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001382 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1383 r = -EINVAL;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001384 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001385 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001386
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001387 kvm_for_each_vcpu(r, v, kvm)
1388 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001389 r = -EEXIST;
1390 goto vcpu_destroy;
1391 }
1392
1393 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001394
1395 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001396 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001397 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001398 if (r < 0) {
1399 kvm_put_kvm(kvm);
1400 goto vcpu_destroy;
1401 }
1402
1403 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1404 smp_wmb();
1405 atomic_inc(&kvm->online_vcpus);
1406
1407#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1408 if (kvm->bsp_vcpu_id == id)
1409 kvm->bsp_vcpu = vcpu;
1410#endif
1411 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001412 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001413
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001414vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001415 mutex_unlock(&kvm->lock);
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001416 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001417 return r;
1418}
1419
Avi Kivity1961d272007-03-05 19:46:05 +02001420static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1421{
1422 if (sigset) {
1423 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1424 vcpu->sigset_active = 1;
1425 vcpu->sigset = *sigset;
1426 } else
1427 vcpu->sigset_active = 0;
1428 return 0;
1429}
1430
Avi Kivitybccf2152007-02-21 18:04:26 +02001431static long kvm_vcpu_ioctl(struct file *filp,
1432 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001433{
Avi Kivitybccf2152007-02-21 18:04:26 +02001434 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +00001435 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001436 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001437 struct kvm_fpu *fpu = NULL;
1438 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001439
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001440 if (vcpu->kvm->mm != current->mm)
1441 return -EIO;
Avi Kivity2122ff52010-05-13 11:25:04 +03001442
1443#if defined(CONFIG_S390) || defined(CONFIG_PPC)
1444 /*
1445 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
1446 * so vcpu_load() would break it.
1447 */
1448 if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT)
1449 return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
1450#endif
1451
1452
1453 vcpu_load(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001454 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001455 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001456 r = -EINVAL;
1457 if (arg)
1458 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001459 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001460 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001461 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001462 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001463
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001464 r = -ENOMEM;
1465 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1466 if (!kvm_regs)
1467 goto out;
1468 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001469 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001470 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001471 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001472 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1473 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001474 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001475out_free1:
1476 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001477 break;
1478 }
1479 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001480 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001481
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001482 r = -ENOMEM;
1483 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1484 if (!kvm_regs)
1485 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001486 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001487 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1488 goto out_free2;
1489 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001490 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001491 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001492 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001493out_free2:
1494 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001495 break;
1496 }
1497 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001498 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1499 r = -ENOMEM;
1500 if (!kvm_sregs)
1501 goto out;
1502 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001503 if (r)
1504 goto out;
1505 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001506 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001507 goto out;
1508 r = 0;
1509 break;
1510 }
1511 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001512 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1513 r = -ENOMEM;
1514 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001515 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001516 r = -EFAULT;
1517 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1518 goto out;
1519 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001520 if (r)
1521 goto out;
1522 r = 0;
1523 break;
1524 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001525 case KVM_GET_MP_STATE: {
1526 struct kvm_mp_state mp_state;
1527
1528 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1529 if (r)
1530 goto out;
1531 r = -EFAULT;
1532 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1533 goto out;
1534 r = 0;
1535 break;
1536 }
1537 case KVM_SET_MP_STATE: {
1538 struct kvm_mp_state mp_state;
1539
1540 r = -EFAULT;
1541 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1542 goto out;
1543 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1544 if (r)
1545 goto out;
1546 r = 0;
1547 break;
1548 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001549 case KVM_TRANSLATE: {
1550 struct kvm_translation tr;
1551
1552 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001553 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001554 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001555 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001556 if (r)
1557 goto out;
1558 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001559 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001560 goto out;
1561 r = 0;
1562 break;
1563 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001564 case KVM_SET_GUEST_DEBUG: {
1565 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001566
1567 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001568 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001569 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001570 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001571 if (r)
1572 goto out;
1573 r = 0;
1574 break;
1575 }
Avi Kivity1961d272007-03-05 19:46:05 +02001576 case KVM_SET_SIGNAL_MASK: {
1577 struct kvm_signal_mask __user *sigmask_arg = argp;
1578 struct kvm_signal_mask kvm_sigmask;
1579 sigset_t sigset, *p;
1580
1581 p = NULL;
1582 if (argp) {
1583 r = -EFAULT;
1584 if (copy_from_user(&kvm_sigmask, argp,
1585 sizeof kvm_sigmask))
1586 goto out;
1587 r = -EINVAL;
1588 if (kvm_sigmask.len != sizeof sigset)
1589 goto out;
1590 r = -EFAULT;
1591 if (copy_from_user(&sigset, sigmask_arg->sigset,
1592 sizeof sigset))
1593 goto out;
1594 p = &sigset;
1595 }
Andi Kleen376d41f2010-06-10 13:10:47 +02001596 r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
Avi Kivity1961d272007-03-05 19:46:05 +02001597 break;
1598 }
Avi Kivityb8836732007-04-01 16:34:31 +03001599 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001600 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1601 r = -ENOMEM;
1602 if (!fpu)
1603 goto out;
1604 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001605 if (r)
1606 goto out;
1607 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001608 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001609 goto out;
1610 r = 0;
1611 break;
1612 }
1613 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001614 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1615 r = -ENOMEM;
1616 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03001617 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001618 r = -EFAULT;
1619 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
1620 goto out;
1621 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001622 if (r)
1623 goto out;
1624 r = 0;
1625 break;
1626 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001627 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001628 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001629 }
1630out:
Avi Kivity2122ff52010-05-13 11:25:04 +03001631 vcpu_put(vcpu);
Dave Hansenfa3795a2008-08-11 10:01:46 -07001632 kfree(fpu);
1633 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001634 return r;
1635}
1636
1637static long kvm_vm_ioctl(struct file *filp,
1638 unsigned int ioctl, unsigned long arg)
1639{
1640 struct kvm *kvm = filp->private_data;
1641 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001642 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001643
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001644 if (kvm->mm != current->mm)
1645 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001646 switch (ioctl) {
1647 case KVM_CREATE_VCPU:
1648 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1649 if (r < 0)
1650 goto out;
1651 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001652 case KVM_SET_USER_MEMORY_REGION: {
1653 struct kvm_userspace_memory_region kvm_userspace_mem;
1654
1655 r = -EFAULT;
1656 if (copy_from_user(&kvm_userspace_mem, argp,
1657 sizeof kvm_userspace_mem))
1658 goto out;
1659
1660 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001661 if (r)
1662 goto out;
1663 break;
1664 }
1665 case KVM_GET_DIRTY_LOG: {
1666 struct kvm_dirty_log log;
1667
1668 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001669 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001670 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02001671 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001672 if (r)
1673 goto out;
1674 break;
1675 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02001676#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1677 case KVM_REGISTER_COALESCED_MMIO: {
1678 struct kvm_coalesced_mmio_zone zone;
1679 r = -EFAULT;
1680 if (copy_from_user(&zone, argp, sizeof zone))
1681 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02001682 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
1683 if (r)
1684 goto out;
1685 r = 0;
1686 break;
1687 }
1688 case KVM_UNREGISTER_COALESCED_MMIO: {
1689 struct kvm_coalesced_mmio_zone zone;
1690 r = -EFAULT;
1691 if (copy_from_user(&zone, argp, sizeof zone))
1692 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02001693 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
1694 if (r)
1695 goto out;
1696 r = 0;
1697 break;
1698 }
1699#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04001700 case KVM_IRQFD: {
1701 struct kvm_irqfd data;
1702
1703 r = -EFAULT;
1704 if (copy_from_user(&data, argp, sizeof data))
1705 goto out;
1706 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
1707 break;
1708 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04001709 case KVM_IOEVENTFD: {
1710 struct kvm_ioeventfd data;
1711
1712 r = -EFAULT;
1713 if (copy_from_user(&data, argp, sizeof data))
1714 goto out;
1715 r = kvm_ioeventfd(kvm, &data);
1716 break;
1717 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001718#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1719 case KVM_SET_BOOT_CPU_ID:
1720 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03001721 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001722 if (atomic_read(&kvm->online_vcpus) != 0)
1723 r = -EBUSY;
1724 else
1725 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03001726 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001727 break;
1728#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001729 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01001730 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03001731 if (r == -ENOTTY)
1732 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02001733 }
1734out:
1735 return r;
1736}
1737
Arnd Bergmann6ff58942009-10-22 14:19:27 +02001738#ifdef CONFIG_COMPAT
1739struct compat_kvm_dirty_log {
1740 __u32 slot;
1741 __u32 padding1;
1742 union {
1743 compat_uptr_t dirty_bitmap; /* one bit per page */
1744 __u64 padding2;
1745 };
1746};
1747
1748static long kvm_vm_compat_ioctl(struct file *filp,
1749 unsigned int ioctl, unsigned long arg)
1750{
1751 struct kvm *kvm = filp->private_data;
1752 int r;
1753
1754 if (kvm->mm != current->mm)
1755 return -EIO;
1756 switch (ioctl) {
1757 case KVM_GET_DIRTY_LOG: {
1758 struct compat_kvm_dirty_log compat_log;
1759 struct kvm_dirty_log log;
1760
1761 r = -EFAULT;
1762 if (copy_from_user(&compat_log, (void __user *)arg,
1763 sizeof(compat_log)))
1764 goto out;
1765 log.slot = compat_log.slot;
1766 log.padding1 = compat_log.padding1;
1767 log.padding2 = compat_log.padding2;
1768 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
1769
1770 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
1771 if (r)
1772 goto out;
1773 break;
1774 }
1775 default:
1776 r = kvm_vm_ioctl(filp, ioctl, arg);
1777 }
1778
1779out:
1780 return r;
1781}
1782#endif
1783
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001784static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02001785{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001786 struct page *page[1];
1787 unsigned long addr;
1788 int npages;
1789 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02001790 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02001791
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001792 addr = gfn_to_hva(kvm, gfn);
1793 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001794 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001795
1796 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
1797 NULL);
1798 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001799 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001800
1801 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001802 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02001803}
1804
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001805static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001806 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02001807};
1808
1809static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
1810{
1811 vma->vm_ops = &kvm_vm_vm_ops;
1812 return 0;
1813}
1814
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001815static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02001816 .release = kvm_vm_release,
1817 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02001818#ifdef CONFIG_COMPAT
1819 .compat_ioctl = kvm_vm_compat_ioctl,
1820#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001821 .mmap = kvm_vm_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001822 .llseek = noop_llseek,
Avi Kivityf17abe92007-02-21 19:28:04 +02001823};
1824
1825static int kvm_dev_ioctl_create_vm(void)
1826{
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09001827 int fd, r;
Avi Kivityf17abe92007-02-21 19:28:04 +02001828 struct kvm *kvm;
1829
Avi Kivityf17abe92007-02-21 19:28:04 +02001830 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04001831 if (IS_ERR(kvm))
1832 return PTR_ERR(kvm);
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09001833#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1834 r = kvm_coalesced_mmio_init(kvm);
1835 if (r < 0) {
1836 kvm_put_kvm(kvm);
1837 return r;
1838 }
1839#endif
Roland Dreier628ff7c2009-12-18 09:41:24 -08001840 fd = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
Al Viro2030a422008-02-23 06:46:49 -05001841 if (fd < 0)
Al Viro66c0b392008-04-19 20:33:56 +01001842 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02001843
Avi Kivityf17abe92007-02-21 19:28:04 +02001844 return fd;
Avi Kivityf17abe92007-02-21 19:28:04 +02001845}
1846
Avi Kivity1a811b62008-12-08 18:25:27 +02001847static long kvm_dev_ioctl_check_extension_generic(long arg)
1848{
1849 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02001850 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02001851 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02001852 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03001853#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1854 case KVM_CAP_SET_BOOT_CPU_ID:
1855#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02001856 case KVM_CAP_INTERNAL_ERROR_DATA:
Avi Kivity1a811b62008-12-08 18:25:27 +02001857 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02001858#ifdef CONFIG_HAVE_KVM_IRQCHIP
1859 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08001860 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02001861#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02001862 default:
1863 break;
1864 }
1865 return kvm_dev_ioctl_check_extension(arg);
1866}
1867
Avi Kivityf17abe92007-02-21 19:28:04 +02001868static long kvm_dev_ioctl(struct file *filp,
1869 unsigned int ioctl, unsigned long arg)
1870{
Avi Kivity07c45a32007-03-07 13:05:38 +02001871 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02001872
1873 switch (ioctl) {
1874 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001875 r = -EINVAL;
1876 if (arg)
1877 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02001878 r = KVM_API_VERSION;
1879 break;
1880 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001881 r = -EINVAL;
1882 if (arg)
1883 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02001884 r = kvm_dev_ioctl_create_vm();
1885 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08001886 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02001887 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02001888 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02001889 case KVM_GET_VCPU_MMAP_SIZE:
1890 r = -EINVAL;
1891 if (arg)
1892 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02001893 r = PAGE_SIZE; /* struct kvm_run */
1894#ifdef CONFIG_X86
1895 r += PAGE_SIZE; /* pio data page */
1896#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001897#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1898 r += PAGE_SIZE; /* coalesced mmio ring page */
1899#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02001900 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04001901 case KVM_TRACE_ENABLE:
1902 case KVM_TRACE_PAUSE:
1903 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03001904 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04001905 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001906 default:
Carsten Otte043405e2007-10-10 17:16:19 +02001907 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001908 }
1909out:
1910 return r;
1911}
1912
Avi Kivity6aa8b732006-12-10 02:21:36 -08001913static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08001914 .unlocked_ioctl = kvm_dev_ioctl,
1915 .compat_ioctl = kvm_dev_ioctl,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001916 .llseek = noop_llseek,
Avi Kivity6aa8b732006-12-10 02:21:36 -08001917};
1918
1919static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02001920 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08001921 "kvm",
1922 &kvm_chardev_ops,
1923};
1924
Avi Kivity1b6c0162007-05-24 13:03:52 +03001925static void hardware_enable(void *junk)
1926{
1927 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02001928 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03001929
Rusty Russell7f59f492008-12-07 21:25:45 +10301930 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03001931 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02001932
Rusty Russell7f59f492008-12-07 21:25:45 +10301933 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02001934
1935 r = kvm_arch_hardware_enable(NULL);
1936
1937 if (r) {
1938 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
1939 atomic_inc(&hardware_enable_failed);
1940 printk(KERN_INFO "kvm: enabling virtualization on "
1941 "CPU%d failed\n", cpu);
1942 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03001943}
1944
1945static void hardware_disable(void *junk)
1946{
1947 int cpu = raw_smp_processor_id();
1948
Rusty Russell7f59f492008-12-07 21:25:45 +10301949 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03001950 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10301951 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001952 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03001953}
1954
Alexander Graf10474ae2009-09-15 11:37:46 +02001955static void hardware_disable_all_nolock(void)
1956{
1957 BUG_ON(!kvm_usage_count);
1958
1959 kvm_usage_count--;
1960 if (!kvm_usage_count)
1961 on_each_cpu(hardware_disable, NULL, 1);
1962}
1963
1964static void hardware_disable_all(void)
1965{
1966 spin_lock(&kvm_lock);
1967 hardware_disable_all_nolock();
1968 spin_unlock(&kvm_lock);
1969}
1970
1971static int hardware_enable_all(void)
1972{
1973 int r = 0;
1974
1975 spin_lock(&kvm_lock);
1976
1977 kvm_usage_count++;
1978 if (kvm_usage_count == 1) {
1979 atomic_set(&hardware_enable_failed, 0);
1980 on_each_cpu(hardware_enable, NULL, 1);
1981
1982 if (atomic_read(&hardware_enable_failed)) {
1983 hardware_disable_all_nolock();
1984 r = -EBUSY;
1985 }
1986 }
1987
1988 spin_unlock(&kvm_lock);
1989
1990 return r;
1991}
1992
Avi Kivity774c47f2007-02-12 00:54:47 -08001993static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
1994 void *v)
1995{
1996 int cpu = (long)v;
1997
Alexander Graf10474ae2009-09-15 11:37:46 +02001998 if (!kvm_usage_count)
1999 return NOTIFY_OK;
2000
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002001 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002002 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002003 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002004 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2005 cpu);
2006 hardware_disable(NULL);
2007 break;
Zachary Amsdenda908f22010-08-19 22:07:27 -10002008 case CPU_STARTING:
Jeremy Katz43934a32007-02-19 14:37:46 +02002009 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2010 cpu);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002011 spin_lock(&kvm_lock);
Zachary Amsdenda908f22010-08-19 22:07:27 -10002012 hardware_enable(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002013 spin_unlock(&kvm_lock);
Avi Kivity774c47f2007-02-12 00:54:47 -08002014 break;
2015 }
2016 return NOTIFY_OK;
2017}
2018
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002019
2020asmlinkage void kvm_handle_fault_on_reboot(void)
2021{
Avi Kivityca242ac2010-09-21 19:59:43 +02002022 if (kvm_rebooting) {
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002023 /* spin while reset goes on */
Avi Kivityca242ac2010-09-21 19:59:43 +02002024 local_irq_enable();
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002025 while (true)
Avi Kivity624d84c2010-09-21 19:59:44 +02002026 cpu_relax();
Avi Kivityca242ac2010-09-21 19:59:43 +02002027 }
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002028 /* Fault while not rebooting. We want the trace. */
2029 BUG();
2030}
2031EXPORT_SYMBOL_GPL(kvm_handle_fault_on_reboot);
2032
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002033static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002034 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002035{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002036 /*
2037 * Some (well, at least mine) BIOSes hang on reboot if
2038 * in vmx root mode.
2039 *
2040 * And Intel TXT required VMX off for all cpu when system shutdown.
2041 */
2042 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2043 kvm_rebooting = true;
2044 on_each_cpu(hardware_disable, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002045 return NOTIFY_OK;
2046}
2047
2048static struct notifier_block kvm_reboot_notifier = {
2049 .notifier_call = kvm_reboot,
2050 .priority = 0,
2051};
2052
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002053static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002054{
2055 int i;
2056
2057 for (i = 0; i < bus->dev_count; i++) {
2058 struct kvm_io_device *pos = bus->devs[i];
2059
2060 kvm_iodevice_destructor(pos);
2061 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002062 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002063}
2064
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002065/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002066int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002067 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002068{
2069 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002070 struct kvm_io_bus *bus;
2071
2072 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002073 for (i = 0; i < bus->dev_count; i++)
2074 if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
2075 return 0;
2076 return -EOPNOTSUPP;
2077}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002078
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002079/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002080int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2081 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002082{
2083 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002084 struct kvm_io_bus *bus;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002085
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002086 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002087 for (i = 0; i < bus->dev_count; i++)
2088 if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
2089 return 0;
2090 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002091}
2092
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002093/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002094int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2095 struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002096{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002097 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002098
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002099 bus = kvm->buses[bus_idx];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002100 if (bus->dev_count > NR_IOBUS_DEVS-1)
2101 return -ENOSPC;
2102
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002103 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2104 if (!new_bus)
2105 return -ENOMEM;
2106 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2107 new_bus->devs[new_bus->dev_count++] = dev;
2108 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2109 synchronize_srcu_expedited(&kvm->srcu);
2110 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002111
2112 return 0;
2113}
2114
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002115/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002116int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2117 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002118{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002119 int i, r;
2120 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002121
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002122 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2123 if (!new_bus)
2124 return -ENOMEM;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002125
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002126 bus = kvm->buses[bus_idx];
2127 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2128
2129 r = -ENOENT;
2130 for (i = 0; i < new_bus->dev_count; i++)
2131 if (new_bus->devs[i] == dev) {
2132 r = 0;
2133 new_bus->devs[i] = new_bus->devs[--new_bus->dev_count];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002134 break;
2135 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002136
2137 if (r) {
2138 kfree(new_bus);
2139 return r;
2140 }
2141
2142 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2143 synchronize_srcu_expedited(&kvm->srcu);
2144 kfree(bus);
2145 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002146}
2147
Avi Kivity774c47f2007-02-12 00:54:47 -08002148static struct notifier_block kvm_cpu_notifier = {
2149 .notifier_call = kvm_cpu_hotplug,
Avi Kivity774c47f2007-02-12 00:54:47 -08002150};
2151
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002152static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002153{
2154 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002155 struct kvm *kvm;
2156
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002157 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002158 spin_lock(&kvm_lock);
2159 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002160 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02002161 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002162 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002163}
2164
2165DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2166
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002167static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002168{
2169 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002170 struct kvm *kvm;
2171 struct kvm_vcpu *vcpu;
2172 int i;
2173
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002174 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002175 spin_lock(&kvm_lock);
2176 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002177 kvm_for_each_vcpu(i, vcpu, kvm)
2178 *val += *(u32 *)((void *)vcpu + offset);
2179
Avi Kivity1165f5f2007-04-19 17:27:43 +03002180 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002181 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002182}
2183
Avi Kivityba1389b2007-11-18 16:24:12 +02002184DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2185
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002186static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002187 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2188 [KVM_STAT_VM] = &vm_stat_fops,
2189};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002190
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002191static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002192{
2193 struct kvm_stats_debugfs_item *p;
2194
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002195 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002196 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002197 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002198 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002199 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002200}
2201
2202static void kvm_exit_debug(void)
2203{
2204 struct kvm_stats_debugfs_item *p;
2205
2206 for (p = debugfs_entries; p->name; ++p)
2207 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002208 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002209}
2210
Avi Kivity59ae6c62007-02-12 00:54:48 -08002211static int kvm_suspend(struct sys_device *dev, pm_message_t state)
2212{
Alexander Graf10474ae2009-09-15 11:37:46 +02002213 if (kvm_usage_count)
2214 hardware_disable(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002215 return 0;
2216}
2217
2218static int kvm_resume(struct sys_device *dev)
2219{
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002220 if (kvm_usage_count) {
2221 WARN_ON(spin_is_locked(&kvm_lock));
Alexander Graf10474ae2009-09-15 11:37:46 +02002222 hardware_enable(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002223 }
Avi Kivity59ae6c62007-02-12 00:54:48 -08002224 return 0;
2225}
2226
2227static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01002228 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08002229 .suspend = kvm_suspend,
2230 .resume = kvm_resume,
2231};
2232
2233static struct sys_device kvm_sysdev = {
2234 .id = 0,
2235 .cls = &kvm_sysdev_class,
2236};
2237
Izik Eiduscea7bb22007-10-17 19:17:48 +02002238struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002239pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002240
Avi Kivity15ad7142007-07-11 18:17:21 +03002241static inline
2242struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2243{
2244 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2245}
2246
2247static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2248{
2249 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2250
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002251 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002252}
2253
2254static void kvm_sched_out(struct preempt_notifier *pn,
2255 struct task_struct *next)
2256{
2257 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2258
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002259 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002260}
2261
Avi Kivity0ee75be2010-04-28 15:39:01 +03002262int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
Rusty Russellc16f8622007-07-30 21:12:19 +10002263 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002264{
2265 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002266 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002267
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002268 r = kvm_arch_init(opaque);
2269 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002270 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002271
2272 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2273
2274 if (bad_page == NULL) {
2275 r = -ENOMEM;
2276 goto out;
2277 }
2278
Anthony Liguori35149e22008-04-02 14:46:56 -05002279 bad_pfn = page_to_pfn(bad_page);
2280
Huang Yingbf998152010-05-31 14:28:19 +08002281 hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2282
2283 if (hwpoison_page == NULL) {
2284 r = -ENOMEM;
2285 goto out_free_0;
2286 }
2287
2288 hwpoison_pfn = page_to_pfn(hwpoison_page);
2289
Gleb Natapovedba23e2010-07-07 20:16:45 +03002290 fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2291
2292 if (fault_page == NULL) {
2293 r = -ENOMEM;
2294 goto out_free_0;
2295 }
2296
2297 fault_pfn = page_to_pfn(fault_page);
2298
Avi Kivity8437a612009-06-06 14:52:35 -07002299 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302300 r = -ENOMEM;
2301 goto out_free_0;
2302 }
2303
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002304 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002305 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302306 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002307
Yang, Sheng002c7f72007-07-31 14:23:01 +03002308 for_each_online_cpu(cpu) {
2309 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002310 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002311 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002312 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002313 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002314 }
2315
Avi Kivity774c47f2007-02-12 00:54:47 -08002316 r = register_cpu_notifier(&kvm_cpu_notifier);
2317 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002318 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002319 register_reboot_notifier(&kvm_reboot_notifier);
2320
Avi Kivity59ae6c62007-02-12 00:54:48 -08002321 r = sysdev_class_register(&kvm_sysdev_class);
2322 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002323 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002324
2325 r = sysdev_register(&kvm_sysdev);
2326 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002327 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002328
Rusty Russellc16f8622007-07-30 21:12:19 +10002329 /* A kmem cache lets us meet the alignment requirements of fx_save. */
Avi Kivity0ee75be2010-04-28 15:39:01 +03002330 if (!vcpu_align)
2331 vcpu_align = __alignof__(struct kvm_vcpu);
2332 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
Joe Perches56919c52007-11-12 20:06:51 -08002333 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002334 if (!kvm_vcpu_cache) {
2335 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08002336 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10002337 }
2338
Avi Kivity6aa8b732006-12-10 02:21:36 -08002339 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002340 kvm_vm_fops.owner = module;
2341 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002342
2343 r = misc_register(&kvm_dev);
2344 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002345 printk(KERN_ERR "kvm: misc device register failed\n");
Avi Kivity6aa8b732006-12-10 02:21:36 -08002346 goto out_free;
2347 }
2348
Avi Kivity15ad7142007-07-11 18:17:21 +03002349 kvm_preempt_ops.sched_in = kvm_sched_in;
2350 kvm_preempt_ops.sched_out = kvm_sched_out;
2351
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002352 kvm_init_debug();
2353
Avi Kivityc7addb92007-09-16 18:58:32 +02002354 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002355
2356out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002357 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002358out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002359 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002360out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002361 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002362out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002363 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002364 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002365out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002366out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002367 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302368out_free_0a:
2369 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002370out_free_0:
Gleb Natapovedba23e2010-07-07 20:16:45 +03002371 if (fault_page)
2372 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002373 if (hwpoison_page)
2374 __free_page(hwpoison_page);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002375 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002376out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002377 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002378out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002379 return r;
2380}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002381EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002382
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002383void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002384{
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002385 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002386 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002387 kmem_cache_destroy(kvm_vcpu_cache);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002388 sysdev_unregister(&kvm_sysdev);
2389 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002390 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002391 unregister_cpu_notifier(&kvm_cpu_notifier);
Jens Axboe15c8b6c2008-05-09 09:39:44 +02002392 on_each_cpu(hardware_disable, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002393 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002394 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302395 free_cpumask_var(cpus_hardware_enabled);
Huang Yingbf998152010-05-31 14:28:19 +08002396 __free_page(hwpoison_page);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002397 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002398}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002399EXPORT_SYMBOL_GPL(kvm_exit);