blob: f29abeb6a9121782874aa9968dbfcdadf16afe42 [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"
Gleb Natapovaf585b92010-10-14 11:22:46 +020058#include "async_pf.h"
Laurent Vivier5f94c172008-05-30 16:05:54 +020059
Marcelo Tosatti229456f2009-06-17 09:22:14 -030060#define CREATE_TRACE_POINTS
61#include <trace/events/kvm.h>
62
Avi Kivity6aa8b732006-12-10 02:21:36 -080063MODULE_AUTHOR("Qumranet");
64MODULE_LICENSE("GPL");
65
Marcelo Tosattifa40a822009-06-04 15:08:24 -030066/*
67 * Ordering of locks:
68 *
Sheng Yangfae3a352009-12-15 10:28:07 +080069 * kvm->lock --> kvm->slots_lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030070 */
71
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080072DEFINE_SPINLOCK(kvm_lock);
73LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080074
Rusty Russell7f59f492008-12-07 21:25:45 +103075static cpumask_var_t cpus_hardware_enabled;
Alexander Graf10474ae2009-09-15 11:37:46 +020076static int kvm_usage_count = 0;
77static atomic_t hardware_enable_failed;
Avi Kivity1b6c0162007-05-24 13:03:52 +030078
Rusty Russellc16f8622007-07-30 21:12:19 +100079struct kmem_cache *kvm_vcpu_cache;
80EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030081
Avi Kivity15ad7142007-07-11 18:17:21 +030082static __read_mostly struct preempt_ops kvm_preempt_ops;
83
Hollis Blanchard76f7c872008-04-15 16:05:42 -050084struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080085
Avi Kivitybccf2152007-02-21 18:04:26 +020086static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
87 unsigned long arg);
Alexander Graf10474ae2009-09-15 11:37:46 +020088static int hardware_enable_all(void);
89static void hardware_disable_all(void);
Avi Kivitybccf2152007-02-21 18:04:26 +020090
Marcelo Tosattie93f8a02009-12-23 14:35:24 -020091static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
92
Avi Kivityb7c41452010-12-02 17:52:50 +020093bool kvm_rebooting;
94EXPORT_SYMBOL_GPL(kvm_rebooting);
Avi Kivity4ecac3f2008-05-13 13:23:38 +030095
Marcelo Tosatti54dee992009-06-11 12:07:44 -030096static bool largepages_enabled = true;
97
Gleb Natapovfa7bff82010-07-07 20:16:44 +030098static struct page *hwpoison_page;
99static pfn_t hwpoison_pfn;
Huang Yingbf998152010-05-31 14:28:19 +0800100
Gleb Natapovedba23e2010-07-07 20:16:45 +0300101static struct page *fault_page;
102static pfn_t fault_pfn;
103
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800104inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300105{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100106 if (pfn_valid(pfn)) {
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800107 int reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800108 struct page *tail = pfn_to_page(pfn);
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800109 struct page *head = compound_trans_head(tail);
110 reserved = PageReserved(head);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800111 if (head != tail) {
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800112 /*
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800113 * "head" is not a dangling pointer
114 * (compound_trans_head takes care of that)
115 * but the hugepage may have been splitted
116 * from under us (and we may not hold a
117 * reference count on the head page so it can
118 * be reused before we run PageReferenced), so
119 * we've to check PageTail before returning
120 * what we just read.
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800121 */
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800122 smp_rmb();
123 if (PageTail(tail))
124 return reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800125 }
126 return PageReserved(tail);
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100127 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300128
129 return true;
130}
131
Avi Kivity6aa8b732006-12-10 02:21:36 -0800132/*
133 * Switches to specified vcpu, until a matching vcpu_put()
134 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200135void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800136{
Avi Kivity15ad7142007-07-11 18:17:21 +0300137 int cpu;
138
Avi Kivitybccf2152007-02-21 18:04:26 +0200139 mutex_lock(&vcpu->mutex);
Avi Kivity15ad7142007-07-11 18:17:21 +0300140 cpu = get_cpu();
141 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200142 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300143 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200144}
145
Carsten Otte313a3dc2007-10-11 19:16:52 +0200146void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800147{
Avi Kivity15ad7142007-07-11 18:17:21 +0300148 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200149 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300150 preempt_notifier_unregister(&vcpu->preempt_notifier);
151 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800152 mutex_unlock(&vcpu->mutex);
153}
154
Avi Kivityd9e368d2007-06-07 19:18:30 +0300155static void ack_flush(void *_completed)
156{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300157}
158
Rusty Russell49846892008-12-08 20:26:24 +1030159static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300160{
Avi Kivity597a5f52008-07-20 14:24:22 +0300161 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030162 cpumask_var_t cpus;
163 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300164 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300165
Li Zefan79f55992009-06-15 14:58:26 +0800166 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030167
Avi Kivity70e335e2010-02-18 11:25:22 +0200168 raw_spin_lock(&kvm->requests_lock);
Jan Kiszkae601e3b2009-07-20 11:30:12 +0200169 me = smp_processor_id();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300170 kvm_for_each_vcpu(i, vcpu, kvm) {
Avi Kivitya8eeb042010-05-10 12:34:53 +0300171 if (kvm_make_check_request(req, vcpu))
Avi Kivityd9e368d2007-06-07 19:18:30 +0300172 continue;
173 cpu = vcpu->cpu;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030174 if (cpus != NULL && cpu != -1 && cpu != me)
175 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300176 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030177 if (unlikely(cpus == NULL))
178 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
179 else if (!cpumask_empty(cpus))
180 smp_call_function_many(cpus, ack_flush, NULL, 1);
181 else
182 called = false;
Avi Kivity70e335e2010-02-18 11:25:22 +0200183 raw_spin_unlock(&kvm->requests_lock);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030184 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030185 return called;
186}
187
188void kvm_flush_remote_tlbs(struct kvm *kvm)
189{
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800190 int dirty_count = kvm->tlbs_dirty;
191
192 smp_mb();
Rusty Russell49846892008-12-08 20:26:24 +1030193 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
194 ++kvm->stat.remote_tlb_flush;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800195 cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300196}
197
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500198void kvm_reload_remote_mmus(struct kvm *kvm)
199{
Rusty Russell49846892008-12-08 20:26:24 +1030200 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500201}
202
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000203int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
204{
205 struct page *page;
206 int r;
207
208 mutex_init(&vcpu->mutex);
209 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000210 vcpu->kvm = kvm;
211 vcpu->vcpu_id = id;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300212 init_waitqueue_head(&vcpu->wq);
Gleb Natapovaf585b92010-10-14 11:22:46 +0200213 kvm_async_pf_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000214
215 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
216 if (!page) {
217 r = -ENOMEM;
218 goto fail;
219 }
220 vcpu->run = page_address(page);
221
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800222 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000223 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800224 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000225 return 0;
226
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000227fail_free_run:
228 free_page((unsigned long)vcpu->run);
229fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000230 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000231}
232EXPORT_SYMBOL_GPL(kvm_vcpu_init);
233
234void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
235{
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800236 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000237 free_page((unsigned long)vcpu->run);
238}
239EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
240
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200241#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
242static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
243{
244 return container_of(mn, struct kvm, mmu_notifier);
245}
246
247static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
248 struct mm_struct *mm,
249 unsigned long address)
250{
251 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200252 int need_tlb_flush, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200253
254 /*
255 * When ->invalidate_page runs, the linux pte has been zapped
256 * already but the page is still allocated until
257 * ->invalidate_page returns. So if we increase the sequence
258 * here the kvm page fault will notice if the spte can't be
259 * established because the page is going to be freed. If
260 * instead the kvm page fault establishes the spte before
261 * ->invalidate_page runs, kvm_unmap_hva will release it
262 * before returning.
263 *
264 * The sequence increase only need to be seen at spin_unlock
265 * time, and not at spin_lock time.
266 *
267 * Increasing the sequence after the spin_unlock would be
268 * unsafe because the kvm page fault could then establish the
269 * pte after kvm_unmap_hva returned, without noticing the page
270 * is going to be freed.
271 */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200272 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200273 spin_lock(&kvm->mmu_lock);
274 kvm->mmu_notifier_seq++;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800275 need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200276 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200277 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200278
279 /* we've to flush the tlb before the pages can be freed */
280 if (need_tlb_flush)
281 kvm_flush_remote_tlbs(kvm);
282
283}
284
Izik Eidus3da0dd42009-09-23 21:47:18 +0300285static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
286 struct mm_struct *mm,
287 unsigned long address,
288 pte_t pte)
289{
290 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200291 int idx;
Izik Eidus3da0dd42009-09-23 21:47:18 +0300292
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200293 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300294 spin_lock(&kvm->mmu_lock);
295 kvm->mmu_notifier_seq++;
296 kvm_set_spte_hva(kvm, address, pte);
297 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200298 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300299}
300
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200301static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
302 struct mm_struct *mm,
303 unsigned long start,
304 unsigned long end)
305{
306 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200307 int need_tlb_flush = 0, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200308
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200309 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200310 spin_lock(&kvm->mmu_lock);
311 /*
312 * The count increase must become visible at unlock time as no
313 * spte can be established without taking the mmu_lock and
314 * count is also read inside the mmu_lock critical section.
315 */
316 kvm->mmu_notifier_count++;
317 for (; start < end; start += PAGE_SIZE)
318 need_tlb_flush |= kvm_unmap_hva(kvm, start);
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800319 need_tlb_flush |= kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200320 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200321 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200322
323 /* we've to flush the tlb before the pages can be freed */
324 if (need_tlb_flush)
325 kvm_flush_remote_tlbs(kvm);
326}
327
328static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
329 struct mm_struct *mm,
330 unsigned long start,
331 unsigned long end)
332{
333 struct kvm *kvm = mmu_notifier_to_kvm(mn);
334
335 spin_lock(&kvm->mmu_lock);
336 /*
337 * This sequence increase will notify the kvm page fault that
338 * the page that is going to be mapped in the spte could have
339 * been freed.
340 */
341 kvm->mmu_notifier_seq++;
342 /*
343 * The above sequence increase must be visible before the
344 * below count decrease but both values are read by the kvm
345 * page fault under mmu_lock spinlock so we don't need to add
346 * a smb_wmb() here in between the two.
347 */
348 kvm->mmu_notifier_count--;
349 spin_unlock(&kvm->mmu_lock);
350
351 BUG_ON(kvm->mmu_notifier_count < 0);
352}
353
354static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
355 struct mm_struct *mm,
356 unsigned long address)
357{
358 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200359 int young, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200360
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200361 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200362 spin_lock(&kvm->mmu_lock);
363 young = kvm_age_hva(kvm, address);
364 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200365 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200366
367 if (young)
368 kvm_flush_remote_tlbs(kvm);
369
370 return young;
371}
372
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800373static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
374 struct mm_struct *mm,
375 unsigned long address)
376{
377 struct kvm *kvm = mmu_notifier_to_kvm(mn);
378 int young, idx;
379
380 idx = srcu_read_lock(&kvm->srcu);
381 spin_lock(&kvm->mmu_lock);
382 young = kvm_test_age_hva(kvm, address);
383 spin_unlock(&kvm->mmu_lock);
384 srcu_read_unlock(&kvm->srcu, idx);
385
386 return young;
387}
388
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100389static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
390 struct mm_struct *mm)
391{
392 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800393 int idx;
394
395 idx = srcu_read_lock(&kvm->srcu);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100396 kvm_arch_flush_shadow(kvm);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800397 srcu_read_unlock(&kvm->srcu, idx);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100398}
399
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200400static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
401 .invalidate_page = kvm_mmu_notifier_invalidate_page,
402 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
403 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
404 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800405 .test_young = kvm_mmu_notifier_test_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300406 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100407 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200408};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200409
410static int kvm_init_mmu_notifier(struct kvm *kvm)
411{
412 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
413 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
414}
415
416#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
417
418static int kvm_init_mmu_notifier(struct kvm *kvm)
419{
420 return 0;
421}
422
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200423#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
424
Avi Kivityf17abe92007-02-21 19:28:04 +0200425static struct kvm *kvm_create_vm(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800426{
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100427 int r, i;
428 struct kvm *kvm = kvm_arch_alloc_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800429
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100430 if (!kvm)
431 return ERR_PTR(-ENOMEM);
432
433 r = kvm_arch_init_vm(kvm);
434 if (r)
435 goto out_err_nodisable;
Alexander Graf10474ae2009-09-15 11:37:46 +0200436
437 r = hardware_enable_all();
438 if (r)
439 goto out_err_nodisable;
440
Avi Kivity75858a82009-01-04 17:10:50 +0200441#ifdef CONFIG_HAVE_KVM_IRQCHIP
442 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300443 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200444#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800445
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200446 r = -ENOMEM;
447 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
448 if (!kvm->memslots)
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100449 goto out_err_nosrcu;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200450 if (init_srcu_struct(&kvm->srcu))
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100451 goto out_err_nosrcu;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200452 for (i = 0; i < KVM_NR_BUSES; i++) {
453 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
454 GFP_KERNEL);
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100455 if (!kvm->buses[i])
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200456 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200457 }
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200458
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200459 r = kvm_init_mmu_notifier(kvm);
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100460 if (r)
Avi Kivity283d0c62009-12-20 14:25:19 +0200461 goto out_err;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200462
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200463 kvm->mm = current->mm;
464 atomic_inc(&kvm->mm->mm_count);
Marcelo Tosattiaaee2c92007-12-20 19:18:26 -0500465 spin_lock_init(&kvm->mmu_lock);
Avi Kivity70e335e2010-02-18 11:25:22 +0200466 raw_spin_lock_init(&kvm->requests_lock);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400467 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800468 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300469 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200470 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300471 atomic_set(&kvm->users_count, 1);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000472 spin_lock(&kvm_lock);
473 list_add(&kvm->vm_list, &vm_list);
474 spin_unlock(&kvm_lock);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100475
Avi Kivityf17abe92007-02-21 19:28:04 +0200476 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200477
478out_err:
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100479 cleanup_srcu_struct(&kvm->srcu);
480out_err_nosrcu:
Alexander Graf10474ae2009-09-15 11:37:46 +0200481 hardware_disable_all();
482out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200483 for (i = 0; i < KVM_NR_BUSES; i++)
484 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200485 kfree(kvm->memslots);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100486 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200487 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200488}
489
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900490static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
491{
492 if (!memslot->dirty_bitmap)
493 return;
494
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900495 if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE)
496 vfree(memslot->dirty_bitmap_head);
497 else
498 kfree(memslot->dirty_bitmap_head);
499
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900500 memslot->dirty_bitmap = NULL;
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900501 memslot->dirty_bitmap_head = NULL;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900502}
503
Avi Kivity6aa8b732006-12-10 02:21:36 -0800504/*
505 * Free any memory in @free but not in @dont.
506 */
507static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
508 struct kvm_memory_slot *dont)
509{
Joerg Roedelec04b262009-06-19 15:16:23 +0200510 int i;
511
Izik Eidus290fc382007-09-27 14:11:22 +0200512 if (!dont || free->rmap != dont->rmap)
513 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800514
515 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900516 kvm_destroy_dirty_bitmap(free);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800517
Joerg Roedelec04b262009-06-19 15:16:23 +0200518
519 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
520 if (!dont || free->lpage_info[i] != dont->lpage_info[i]) {
521 vfree(free->lpage_info[i]);
522 free->lpage_info[i] = NULL;
523 }
524 }
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300525
Avi Kivity6aa8b732006-12-10 02:21:36 -0800526 free->npages = 0;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500527 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800528}
529
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800530void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800531{
532 int i;
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200533 struct kvm_memslots *slots = kvm->memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800534
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200535 for (i = 0; i < slots->nmemslots; ++i)
536 kvm_free_physmem_slot(&slots->memslots[i], NULL);
537
538 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800539}
540
Avi Kivityf17abe92007-02-21 19:28:04 +0200541static void kvm_destroy_vm(struct kvm *kvm)
542{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200543 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200544 struct mm_struct *mm = kvm->mm;
545
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800546 kvm_arch_sync_events(kvm);
Avi Kivity133de902007-02-12 00:54:44 -0800547 spin_lock(&kvm_lock);
548 list_del(&kvm->vm_list);
549 spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200550 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200551 for (i = 0; i < KVM_NR_BUSES; i++)
552 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200553 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200554#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
555 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200556#else
557 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200558#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800559 kvm_arch_destroy_vm(kvm);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100560 kvm_free_physmem(kvm);
561 cleanup_srcu_struct(&kvm->srcu);
562 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200563 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200564 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200565}
566
Izik Eidusd39f13b2008-03-30 16:01:25 +0300567void kvm_get_kvm(struct kvm *kvm)
568{
569 atomic_inc(&kvm->users_count);
570}
571EXPORT_SYMBOL_GPL(kvm_get_kvm);
572
573void kvm_put_kvm(struct kvm *kvm)
574{
575 if (atomic_dec_and_test(&kvm->users_count))
576 kvm_destroy_vm(kvm);
577}
578EXPORT_SYMBOL_GPL(kvm_put_kvm);
579
580
Avi Kivityf17abe92007-02-21 19:28:04 +0200581static int kvm_vm_release(struct inode *inode, struct file *filp)
582{
583 struct kvm *kvm = filp->private_data;
584
Gregory Haskins721eecb2009-05-20 10:30:49 -0400585 kvm_irqfd_release(kvm);
586
Izik Eidusd39f13b2008-03-30 16:01:25 +0300587 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800588 return 0;
589}
590
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900591/*
592 * Allocation size is twice as large as the actual dirty bitmap size.
593 * This makes it possible to do double buffering: see x86's
594 * kvm_vm_ioctl_get_dirty_log().
595 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900596static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
597{
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900598 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900599
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900600 if (dirty_bytes > PAGE_SIZE)
601 memslot->dirty_bitmap = vzalloc(dirty_bytes);
602 else
603 memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL);
604
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900605 if (!memslot->dirty_bitmap)
606 return -ENOMEM;
607
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900608 memslot->dirty_bitmap_head = memslot->dirty_bitmap;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900609 return 0;
610}
611
Avi Kivity6aa8b732006-12-10 02:21:36 -0800612/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800613 * Allocate some memory and give it an address in the guest physical address
614 * space.
615 *
616 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800617 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500618 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800619 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800620int __kvm_set_memory_region(struct kvm *kvm,
621 struct kvm_userspace_memory_region *mem,
622 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800623{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200624 int r, flush_shadow = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800625 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200626 unsigned long npages;
627 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800628 struct kvm_memory_slot *memslot;
629 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200630 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800631
632 r = -EINVAL;
633 /* General sanity checks */
634 if (mem->memory_size & (PAGE_SIZE - 1))
635 goto out;
636 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
637 goto out;
Sheng Yange7cacd42008-11-11 15:30:40 +0800638 if (user_alloc && (mem->userspace_addr & (PAGE_SIZE - 1)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600639 goto out;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200640 if (mem->slot >= KVM_MEMORY_SLOTS + KVM_PRIVATE_MEM_SLOTS)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800641 goto out;
642 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
643 goto out;
644
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200645 memslot = &kvm->memslots->memslots[mem->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800646 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
647 npages = mem->memory_size >> PAGE_SHIFT;
648
Takuya Yoshikawa660c22c2010-04-13 22:47:24 +0900649 r = -EINVAL;
650 if (npages > KVM_MEM_MAX_NR_PAGES)
651 goto out;
652
Avi Kivity6aa8b732006-12-10 02:21:36 -0800653 if (!npages)
654 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
655
Avi Kivity6aa8b732006-12-10 02:21:36 -0800656 new = old = *memslot;
657
Avi Kivitye36d96f2010-06-21 10:56:36 +0300658 new.id = mem->slot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800659 new.base_gfn = base_gfn;
660 new.npages = npages;
661 new.flags = mem->flags;
662
663 /* Disallow changing a memory slot's size. */
664 r = -EINVAL;
665 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800666 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800667
668 /* Check for overlaps */
669 r = -EEXIST;
670 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200671 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800672
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200673 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800674 continue;
675 if (!((base_gfn + npages <= s->base_gfn) ||
676 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800677 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800678 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800679
Avi Kivity6aa8b732006-12-10 02:21:36 -0800680 /* Free page dirty bitmap if unneeded */
681 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000682 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800683
684 r = -ENOMEM;
685
686 /* Allocate if a slot is being created */
Carsten Otteeff01142008-06-27 15:05:31 +0200687#ifndef CONFIG_S390
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500688 if (npages && !new.rmap) {
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900689 new.rmap = vzalloc(npages * sizeof(*new.rmap));
Izik Eidus290fc382007-09-27 14:11:22 +0200690
691 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800692 goto out_free;
Izik Eidus290fc382007-09-27 14:11:22 +0200693
Izik Eidus80b14b52007-10-25 11:54:04 +0200694 new.user_alloc = user_alloc;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200695 new.userspace_addr = mem->userspace_addr;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800696 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200697 if (!npages)
698 goto skip_lpage;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300699
Joerg Roedelec04b262009-06-19 15:16:23 +0200700 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
Heiko Carstens28bcb112009-09-03 17:35:35 +0200701 unsigned long ugfn;
702 unsigned long j;
703 int lpages;
Joerg Roedelec04b262009-06-19 15:16:23 +0200704 int level = i + 2;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300705
Joerg Roedelec04b262009-06-19 15:16:23 +0200706 /* Avoid unused variable warning if no large pages */
707 (void)level;
708
709 if (new.lpage_info[i])
710 continue;
711
Joerg Roedel82855412010-07-01 16:00:11 +0200712 lpages = 1 + ((base_gfn + npages - 1)
713 >> KVM_HPAGE_GFN_SHIFT(level));
714 lpages -= base_gfn >> KVM_HPAGE_GFN_SHIFT(level);
Joerg Roedelec04b262009-06-19 15:16:23 +0200715
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900716 new.lpage_info[i] = vzalloc(lpages * sizeof(*new.lpage_info[i]));
Joerg Roedelec04b262009-06-19 15:16:23 +0200717
718 if (!new.lpage_info[i])
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300719 goto out_free;
720
Joerg Roedel82855412010-07-01 16:00:11 +0200721 if (base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200722 new.lpage_info[i][0].write_count = 1;
Joerg Roedel82855412010-07-01 16:00:11 +0200723 if ((base_gfn+npages) & (KVM_PAGES_PER_HPAGE(level) - 1))
Joerg Roedelec04b262009-06-19 15:16:23 +0200724 new.lpage_info[i][lpages - 1].write_count = 1;
Avi Kivityac045272009-06-08 15:52:39 +0300725 ugfn = new.userspace_addr >> PAGE_SHIFT;
726 /*
727 * If the gfn and userspace address are not aligned wrt each
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300728 * other, or if explicitly asked to, disable large page
729 * support for this slot
Avi Kivityac045272009-06-08 15:52:39 +0300730 */
Joerg Roedelec04b262009-06-19 15:16:23 +0200731 if ((base_gfn ^ ugfn) & (KVM_PAGES_PER_HPAGE(level) - 1) ||
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300732 !largepages_enabled)
Joerg Roedelec04b262009-06-19 15:16:23 +0200733 for (j = 0; j < lpages; ++j)
734 new.lpage_info[i][j].write_count = 1;
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300735 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800736
Joerg Roedelec04b262009-06-19 15:16:23 +0200737skip_lpage:
738
Avi Kivity6aa8b732006-12-10 02:21:36 -0800739 /* Allocate page dirty bitmap if needed */
740 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900741 if (kvm_create_dirty_bitmap(&new) < 0)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800742 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200743 /* destroy any largepage mappings for dirty tracking */
Izik Eiduse2445842009-06-10 19:23:24 +0300744 if (old.npages)
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200745 flush_shadow = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800746 }
Christian Borntraeger3eea8432009-06-23 17:24:06 +0200747#else /* not defined CONFIG_S390 */
748 new.user_alloc = user_alloc;
749 if (user_alloc)
750 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200751#endif /* not defined CONFIG_S390 */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800752
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200753 if (!npages) {
754 r = -ENOMEM;
755 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
756 if (!slots)
757 goto out_free;
758 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
759 if (mem->slot >= slots->nmemslots)
760 slots->nmemslots = mem->slot + 1;
Gleb Natapov49c77542010-10-18 15:22:23 +0200761 slots->generation++;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200762 slots->memslots[mem->slot].flags |= KVM_MEMSLOT_INVALID;
763
764 old_memslots = kvm->memslots;
765 rcu_assign_pointer(kvm->memslots, slots);
766 synchronize_srcu_expedited(&kvm->srcu);
767 /* From this point no new shadow pages pointing to a deleted
768 * memslot will be created.
769 *
770 * validation of sp->gfn happens in:
771 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
772 * - kvm_is_visible_gfn (mmu_check_roots)
773 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300774 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200775 kfree(old_memslots);
776 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300777
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200778 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
779 if (r)
780 goto out_free;
781
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200782 /* map the pages in iommu page table */
783 if (npages) {
784 r = kvm_iommu_map_pages(kvm, &new);
785 if (r)
786 goto out_free;
787 }
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200788
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200789 r = -ENOMEM;
790 slots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
791 if (!slots)
792 goto out_free;
793 memcpy(slots, kvm->memslots, sizeof(struct kvm_memslots));
794 if (mem->slot >= slots->nmemslots)
795 slots->nmemslots = mem->slot + 1;
Gleb Natapov49c77542010-10-18 15:22:23 +0200796 slots->generation++;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200797
798 /* actual memory is freed via old in kvm_free_physmem_slot below */
799 if (!npages) {
800 new.rmap = NULL;
801 new.dirty_bitmap = NULL;
802 for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i)
803 new.lpage_info[i] = NULL;
804 }
805
806 slots->memslots[mem->slot] = new;
807 old_memslots = kvm->memslots;
808 rcu_assign_pointer(kvm->memslots, slots);
809 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800810
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200811 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800812
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200813 kvm_free_physmem_slot(&old, &new);
814 kfree(old_memslots);
815
816 if (flush_shadow)
817 kvm_arch_flush_shadow(kvm);
818
Avi Kivity6aa8b732006-12-10 02:21:36 -0800819 return 0;
820
Sheng Yangf78e0e22007-10-29 09:40:42 +0800821out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800822 kvm_free_physmem_slot(&new, &old);
823out:
824 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200825
826}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800827EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
828
829int kvm_set_memory_region(struct kvm *kvm,
830 struct kvm_userspace_memory_region *mem,
831 int user_alloc)
832{
833 int r;
834
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200835 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800836 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200837 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800838 return r;
839}
Izik Eidus210c7c42007-10-24 23:52:57 +0200840EXPORT_SYMBOL_GPL(kvm_set_memory_region);
841
Carsten Otte1fe779f2007-10-29 16:08:35 +0100842int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
843 struct
844 kvm_userspace_memory_region *mem,
845 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200846{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200847 if (mem->slot >= KVM_MEMORY_SLOTS)
848 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200849 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800850}
851
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800852int kvm_get_dirty_log(struct kvm *kvm,
853 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800854{
855 struct kvm_memory_slot *memslot;
856 int r, i;
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900857 unsigned long n;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800858 unsigned long any = 0;
859
Avi Kivity6aa8b732006-12-10 02:21:36 -0800860 r = -EINVAL;
861 if (log->slot >= KVM_MEMORY_SLOTS)
862 goto out;
863
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200864 memslot = &kvm->memslots->memslots[log->slot];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800865 r = -ENOENT;
866 if (!memslot->dirty_bitmap)
867 goto out;
868
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900869 n = kvm_dirty_bitmap_bytes(memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800870
Uri Lublincd1a4a92007-02-22 16:43:09 +0200871 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800872 any = memslot->dirty_bitmap[i];
873
874 r = -EFAULT;
875 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
876 goto out;
877
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800878 if (any)
879 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800880
881 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800882out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800883 return r;
884}
885
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300886void kvm_disable_largepages(void)
887{
888 largepages_enabled = false;
889}
890EXPORT_SYMBOL_GPL(kvm_disable_largepages);
891
Izik Eiduscea7bb22007-10-17 19:17:48 +0200892int is_error_page(struct page *page)
893{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300894 return page == bad_page || page == hwpoison_page || page == fault_page;
Izik Eiduscea7bb22007-10-17 19:17:48 +0200895}
896EXPORT_SYMBOL_GPL(is_error_page);
897
Anthony Liguori35149e22008-04-02 14:46:56 -0500898int is_error_pfn(pfn_t pfn)
899{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300900 return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500901}
902EXPORT_SYMBOL_GPL(is_error_pfn);
903
Huang Yingbf998152010-05-31 14:28:19 +0800904int is_hwpoison_pfn(pfn_t pfn)
905{
906 return pfn == hwpoison_pfn;
907}
908EXPORT_SYMBOL_GPL(is_hwpoison_pfn);
909
Gleb Natapovedba23e2010-07-07 20:16:45 +0300910int is_fault_pfn(pfn_t pfn)
911{
912 return pfn == fault_pfn;
913}
914EXPORT_SYMBOL_GPL(is_fault_pfn);
915
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200916static inline unsigned long bad_hva(void)
917{
918 return PAGE_OFFSET;
919}
920
921int kvm_is_error_hva(unsigned long addr)
922{
923 return addr == bad_hva();
924}
925EXPORT_SYMBOL_GPL(kvm_is_error_hva);
926
Gleb Natapov49c77542010-10-18 15:22:23 +0200927static struct kvm_memory_slot *__gfn_to_memslot(struct kvm_memslots *slots,
928 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800929{
930 int i;
931
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200932 for (i = 0; i < slots->nmemslots; ++i) {
933 struct kvm_memory_slot *memslot = &slots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800934
935 if (gfn >= memslot->base_gfn
936 && gfn < memslot->base_gfn + memslot->npages)
937 return memslot;
938 }
Al Viro8b6d44c2007-02-09 16:38:40 +0000939 return NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800940}
Gleb Natapov49c77542010-10-18 15:22:23 +0200941
942struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
943{
944 return __gfn_to_memslot(kvm_memslots(kvm), gfn);
945}
Avi Kivitya1f4d392010-06-21 11:44:20 +0300946EXPORT_SYMBOL_GPL(gfn_to_memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800947
Izik Eiduse0d62c72007-10-24 23:57:46 +0200948int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
949{
950 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +0800951 struct kvm_memslots *slots = kvm_memslots(kvm);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200952
Izik Eiduse0d62c72007-10-24 23:57:46 +0200953 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200954 struct kvm_memory_slot *memslot = &slots->memslots[i];
Izik Eiduse0d62c72007-10-24 23:57:46 +0200955
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200956 if (memslot->flags & KVM_MEMSLOT_INVALID)
957 continue;
958
Izik Eiduse0d62c72007-10-24 23:57:46 +0200959 if (gfn >= memslot->base_gfn
960 && gfn < memslot->base_gfn + memslot->npages)
961 return 1;
962 }
963 return 0;
964}
965EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
966
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +0100967unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
968{
969 struct vm_area_struct *vma;
970 unsigned long addr, size;
971
972 size = PAGE_SIZE;
973
974 addr = gfn_to_hva(kvm, gfn);
975 if (kvm_is_error_hva(addr))
976 return PAGE_SIZE;
977
978 down_read(&current->mm->mmap_sem);
979 vma = find_vma(current->mm, addr);
980 if (!vma)
981 goto out;
982
983 size = vma_kernel_pagesize(vma);
984
985out:
986 up_read(&current->mm->mmap_sem);
987
988 return size;
989}
990
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200991int memslot_id(struct kvm *kvm, gfn_t gfn)
992{
993 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +0800994 struct kvm_memslots *slots = kvm_memslots(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200995 struct kvm_memory_slot *memslot = NULL;
996
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200997 for (i = 0; i < slots->nmemslots; ++i) {
998 memslot = &slots->memslots[i];
999
1000 if (gfn >= memslot->base_gfn
1001 && gfn < memslot->base_gfn + memslot->npages)
1002 break;
1003 }
1004
1005 return memslot - slots->memslots;
1006}
1007
Gleb Natapov49c77542010-10-18 15:22:23 +02001008static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
Xiao Guangrong48987782010-08-22 19:11:43 +08001009 gfn_t *nr_pages)
Izik Eidus539cb662007-11-11 22:05:04 +02001010{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001011 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +02001012 return bad_hva();
Xiao Guangrong48987782010-08-22 19:11:43 +08001013
1014 if (nr_pages)
1015 *nr_pages = slot->npages - (gfn - slot->base_gfn);
1016
Takuya Yoshikawaf5c98032010-02-25 11:33:19 +09001017 return gfn_to_hva_memslot(slot, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001018}
Xiao Guangrong48987782010-08-22 19:11:43 +08001019
1020unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
1021{
Gleb Natapov49c77542010-10-18 15:22:23 +02001022 return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
Xiao Guangrong48987782010-08-22 19:11:43 +08001023}
Sheng Yang0d150292008-04-25 21:44:50 +08001024EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001025
Gleb Natapov80300892010-10-19 18:13:41 +02001026static pfn_t get_fault_pfn(void)
1027{
1028 get_page(fault_page);
1029 return fault_pfn;
1030}
1031
Gleb Natapovaf585b92010-10-14 11:22:46 +02001032static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic,
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001033 bool *async, bool write_fault, bool *writable)
Avi Kivity954bbbc22007-03-30 14:02:32 +03001034{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001035 struct page *page[1];
Gleb Natapovaf585b92010-10-14 11:22:46 +02001036 int npages = 0;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001037 pfn_t pfn;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001038
Gleb Natapovaf585b92010-10-14 11:22:46 +02001039 /* we can do it either atomically or asynchronously, not both */
1040 BUG_ON(atomic && async);
1041
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001042 BUG_ON(!write_fault && !writable);
1043
1044 if (writable)
1045 *writable = true;
1046
Gleb Natapovaf585b92010-10-14 11:22:46 +02001047 if (atomic || async)
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001048 npages = __get_user_pages_fast(addr, 1, 1, page);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001049
1050 if (unlikely(npages != 1) && !atomic) {
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001051 might_sleep();
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001052
1053 if (writable)
1054 *writable = write_fault;
1055
1056 npages = get_user_pages_fast(addr, 1, write_fault, page);
1057
1058 /* map read fault as writable if possible */
1059 if (unlikely(!write_fault) && npages == 1) {
1060 struct page *wpage[1];
1061
1062 npages = __get_user_pages_fast(addr, 1, 1, wpage);
1063 if (npages == 1) {
1064 *writable = true;
1065 put_page(page[0]);
1066 page[0] = wpage[0];
1067 }
1068 npages = 1;
1069 }
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001070 }
Izik Eidus539cb662007-11-11 22:05:04 +02001071
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001072 if (unlikely(npages != 1)) {
1073 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001074
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001075 if (atomic)
Gleb Natapov80300892010-10-19 18:13:41 +02001076 return get_fault_pfn();
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001077
Huang Yingbbeb3402010-06-22 14:23:11 +08001078 down_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +08001079 if (is_hwpoison_address(addr)) {
Huang Yingbbeb3402010-06-22 14:23:11 +08001080 up_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +08001081 get_page(hwpoison_page);
1082 return page_to_pfn(hwpoison_page);
1083 }
1084
Gleb Natapov80300892010-10-19 18:13:41 +02001085 vma = find_vma_intersection(current->mm, addr, addr+1);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001086
Gleb Natapov80300892010-10-19 18:13:41 +02001087 if (vma == NULL)
1088 pfn = get_fault_pfn();
1089 else if ((vma->vm_flags & VM_PFNMAP)) {
1090 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) +
1091 vma->vm_pgoff;
1092 BUG_ON(!kvm_is_mmio_pfn(pfn));
1093 } else {
1094 if (async && (vma->vm_flags & VM_WRITE))
Gleb Natapovaf585b92010-10-14 11:22:46 +02001095 *async = true;
Gleb Natapov80300892010-10-19 18:13:41 +02001096 pfn = get_fault_pfn();
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001097 }
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001098 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001099 } else
1100 pfn = page_to_pfn(page[0]);
1101
1102 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001103}
1104
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001105pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr)
1106{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001107 return hva_to_pfn(kvm, addr, true, NULL, true, NULL);
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001108}
1109EXPORT_SYMBOL_GPL(hva_to_pfn_atomic);
1110
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001111static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async,
1112 bool write_fault, bool *writable)
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001113{
1114 unsigned long addr;
1115
Gleb Natapovaf585b92010-10-14 11:22:46 +02001116 if (async)
1117 *async = false;
1118
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001119 addr = gfn_to_hva(kvm, gfn);
1120 if (kvm_is_error_hva(addr)) {
1121 get_page(bad_page);
1122 return page_to_pfn(bad_page);
1123 }
1124
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001125 return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001126}
1127
1128pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
1129{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001130 return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001131}
1132EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);
1133
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001134pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
1135 bool write_fault, bool *writable)
Gleb Natapovaf585b92010-10-14 11:22:46 +02001136{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001137 return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001138}
1139EXPORT_SYMBOL_GPL(gfn_to_pfn_async);
1140
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001141pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
1142{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001143 return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001144}
Anthony Liguori35149e22008-04-02 14:46:56 -05001145EXPORT_SYMBOL_GPL(gfn_to_pfn);
1146
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001147pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1148 bool *writable)
1149{
1150 return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable);
1151}
1152EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);
1153
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001154pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
1155 struct kvm_memory_slot *slot, gfn_t gfn)
1156{
1157 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001158 return hva_to_pfn(kvm, addr, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001159}
1160
Xiao Guangrong48987782010-08-22 19:11:43 +08001161int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
1162 int nr_pages)
1163{
1164 unsigned long addr;
1165 gfn_t entry;
1166
Gleb Natapov49c77542010-10-18 15:22:23 +02001167 addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry);
Xiao Guangrong48987782010-08-22 19:11:43 +08001168 if (kvm_is_error_hva(addr))
1169 return -1;
1170
1171 if (entry < nr_pages)
1172 return 0;
1173
1174 return __get_user_pages_fast(addr, nr_pages, 1, pages);
1175}
1176EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);
1177
Anthony Liguori35149e22008-04-02 14:46:56 -05001178struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1179{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001180 pfn_t pfn;
1181
1182 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001183 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001184 return pfn_to_page(pfn);
1185
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001186 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001187
1188 get_page(bad_page);
1189 return bad_page;
Avi Kivity954bbbc22007-03-30 14:02:32 +03001190}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001191
Avi Kivity954bbbc22007-03-30 14:02:32 +03001192EXPORT_SYMBOL_GPL(gfn_to_page);
1193
Izik Eidusb4231d62007-11-20 11:49:33 +02001194void kvm_release_page_clean(struct page *page)
1195{
Anthony Liguori35149e22008-04-02 14:46:56 -05001196 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001197}
1198EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1199
Anthony Liguori35149e22008-04-02 14:46:56 -05001200void kvm_release_pfn_clean(pfn_t pfn)
1201{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001202 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001203 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001204}
1205EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1206
Izik Eidusb4231d62007-11-20 11:49:33 +02001207void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001208{
Anthony Liguori35149e22008-04-02 14:46:56 -05001209 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001210}
Izik Eidusb4231d62007-11-20 11:49:33 +02001211EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001212
Anthony Liguori35149e22008-04-02 14:46:56 -05001213void kvm_release_pfn_dirty(pfn_t pfn)
1214{
1215 kvm_set_pfn_dirty(pfn);
1216 kvm_release_pfn_clean(pfn);
1217}
1218EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1219
1220void kvm_set_page_dirty(struct page *page)
1221{
1222 kvm_set_pfn_dirty(page_to_pfn(page));
1223}
1224EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1225
1226void kvm_set_pfn_dirty(pfn_t pfn)
1227{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001228 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001229 struct page *page = pfn_to_page(pfn);
1230 if (!PageReserved(page))
1231 SetPageDirty(page);
1232 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001233}
1234EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1235
1236void kvm_set_pfn_accessed(pfn_t pfn)
1237{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001238 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001239 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001240}
1241EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1242
1243void kvm_get_pfn(pfn_t pfn)
1244{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001245 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001246 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001247}
1248EXPORT_SYMBOL_GPL(kvm_get_pfn);
1249
Izik Eidus195aefd2007-10-01 22:14:18 +02001250static int next_segment(unsigned long len, int offset)
1251{
1252 if (len > PAGE_SIZE - offset)
1253 return PAGE_SIZE - offset;
1254 else
1255 return len;
1256}
1257
1258int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1259 int len)
1260{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001261 int r;
1262 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001263
Izik Eiduse0506bc2007-11-11 22:10:22 +02001264 addr = gfn_to_hva(kvm, gfn);
1265 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001266 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001267 r = copy_from_user(data, (void __user *)addr + offset, len);
1268 if (r)
1269 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001270 return 0;
1271}
1272EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1273
1274int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1275{
1276 gfn_t gfn = gpa >> PAGE_SHIFT;
1277 int seg;
1278 int offset = offset_in_page(gpa);
1279 int ret;
1280
1281 while ((seg = next_segment(len, offset)) != 0) {
1282 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1283 if (ret < 0)
1284 return ret;
1285 offset = 0;
1286 len -= seg;
1287 data += seg;
1288 ++gfn;
1289 }
1290 return 0;
1291}
1292EXPORT_SYMBOL_GPL(kvm_read_guest);
1293
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001294int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1295 unsigned long len)
1296{
1297 int r;
1298 unsigned long addr;
1299 gfn_t gfn = gpa >> PAGE_SHIFT;
1300 int offset = offset_in_page(gpa);
1301
1302 addr = gfn_to_hva(kvm, gfn);
1303 if (kvm_is_error_hva(addr))
1304 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001305 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001306 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001307 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001308 if (r)
1309 return -EFAULT;
1310 return 0;
1311}
1312EXPORT_SYMBOL(kvm_read_guest_atomic);
1313
Izik Eidus195aefd2007-10-01 22:14:18 +02001314int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1315 int offset, int len)
1316{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001317 int r;
1318 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001319
Izik Eiduse0506bc2007-11-11 22:10:22 +02001320 addr = gfn_to_hva(kvm, gfn);
1321 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001322 return -EFAULT;
Izik Eiduse0506bc2007-11-11 22:10:22 +02001323 r = copy_to_user((void __user *)addr + offset, data, len);
1324 if (r)
1325 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001326 mark_page_dirty(kvm, gfn);
1327 return 0;
1328}
1329EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1330
1331int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1332 unsigned long len)
1333{
1334 gfn_t gfn = gpa >> PAGE_SHIFT;
1335 int seg;
1336 int offset = offset_in_page(gpa);
1337 int ret;
1338
1339 while ((seg = next_segment(len, offset)) != 0) {
1340 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1341 if (ret < 0)
1342 return ret;
1343 offset = 0;
1344 len -= seg;
1345 data += seg;
1346 ++gfn;
1347 }
1348 return 0;
1349}
1350
Gleb Natapov49c77542010-10-18 15:22:23 +02001351int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1352 gpa_t gpa)
1353{
1354 struct kvm_memslots *slots = kvm_memslots(kvm);
1355 int offset = offset_in_page(gpa);
1356 gfn_t gfn = gpa >> PAGE_SHIFT;
1357
1358 ghc->gpa = gpa;
1359 ghc->generation = slots->generation;
1360 ghc->memslot = __gfn_to_memslot(slots, gfn);
1361 ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL);
1362 if (!kvm_is_error_hva(ghc->hva))
1363 ghc->hva += offset;
1364 else
1365 return -EFAULT;
1366
1367 return 0;
1368}
1369EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1370
1371int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1372 void *data, unsigned long len)
1373{
1374 struct kvm_memslots *slots = kvm_memslots(kvm);
1375 int r;
1376
1377 if (slots->generation != ghc->generation)
1378 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1379
1380 if (kvm_is_error_hva(ghc->hva))
1381 return -EFAULT;
1382
1383 r = copy_to_user((void __user *)ghc->hva, data, len);
1384 if (r)
1385 return -EFAULT;
1386 mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT);
1387
1388 return 0;
1389}
1390EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
1391
Izik Eidus195aefd2007-10-01 22:14:18 +02001392int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1393{
Heiko Carstens3bcc8a82010-10-27 17:21:21 +02001394 return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page,
1395 offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001396}
1397EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1398
1399int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1400{
1401 gfn_t gfn = gpa >> PAGE_SHIFT;
1402 int seg;
1403 int offset = offset_in_page(gpa);
1404 int ret;
1405
1406 while ((seg = next_segment(len, offset)) != 0) {
1407 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1408 if (ret < 0)
1409 return ret;
1410 offset = 0;
1411 len -= seg;
1412 ++gfn;
1413 }
1414 return 0;
1415}
1416EXPORT_SYMBOL_GPL(kvm_clear_guest);
1417
Gleb Natapov49c77542010-10-18 15:22:23 +02001418void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
1419 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001420{
Rusty Russell7e9d6192007-07-31 20:41:14 +10001421 if (memslot && memslot->dirty_bitmap) {
1422 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001423
Takuya Yoshikawad1476932010-04-23 17:48:35 +09001424 generic___set_le_bit(rel_gfn, memslot->dirty_bitmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001425 }
1426}
1427
Gleb Natapov49c77542010-10-18 15:22:23 +02001428void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1429{
1430 struct kvm_memory_slot *memslot;
1431
1432 memslot = gfn_to_memslot(kvm, gfn);
1433 mark_page_dirty_in_slot(kvm, memslot, gfn);
1434}
1435
Eddie Dongb6958ce2007-07-18 12:15:21 +03001436/*
1437 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1438 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001439void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001440{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001441 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001442
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001443 for (;;) {
1444 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001445
Gleb Natapova1b37102009-07-09 15:33:52 +03001446 if (kvm_arch_vcpu_runnable(vcpu)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001447 kvm_make_request(KVM_REQ_UNHALT, vcpu);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001448 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001449 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001450 if (kvm_cpu_has_pending_timer(vcpu))
1451 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001452 if (signal_pending(current))
1453 break;
1454
Eddie Dongb6958ce2007-07-18 12:15:21 +03001455 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001456 }
1457
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001458 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001459}
1460
Avi Kivity6aa8b732006-12-10 02:21:36 -08001461void kvm_resched(struct kvm_vcpu *vcpu)
1462{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001463 if (!need_resched())
1464 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001465 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001466}
1467EXPORT_SYMBOL_GPL(kvm_resched);
1468
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001469void kvm_vcpu_on_spin(struct kvm_vcpu *vcpu)
1470{
1471 ktime_t expires;
1472 DEFINE_WAIT(wait);
1473
1474 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
1475
1476 /* Sleep for 100 us, and hope lock-holder got scheduled */
1477 expires = ktime_add_ns(ktime_get(), 100000UL);
1478 schedule_hrtimeout(&expires, HRTIMER_MODE_ABS);
1479
1480 finish_wait(&vcpu->wq, &wait);
1481}
1482EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1483
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001484static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001485{
1486 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001487 struct page *page;
1488
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001489 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001490 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001491#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001492 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001493 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001494#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001495#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1496 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1497 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1498#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001499 else
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001500 return VM_FAULT_SIGBUS;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001501 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001502 vmf->page = page;
1503 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001504}
1505
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001506static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001507 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001508};
1509
1510static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1511{
1512 vma->vm_ops = &kvm_vcpu_vm_ops;
1513 return 0;
1514}
1515
Avi Kivitybccf2152007-02-21 18:04:26 +02001516static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1517{
1518 struct kvm_vcpu *vcpu = filp->private_data;
1519
Al Viro66c0b392008-04-19 20:33:56 +01001520 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001521 return 0;
1522}
1523
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001524static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001525 .release = kvm_vcpu_release,
1526 .unlocked_ioctl = kvm_vcpu_ioctl,
1527 .compat_ioctl = kvm_vcpu_ioctl,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001528 .mmap = kvm_vcpu_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001529 .llseek = noop_llseek,
Avi Kivitybccf2152007-02-21 18:04:26 +02001530};
1531
1532/*
1533 * Allocates an inode for the vcpu.
1534 */
1535static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1536{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001537 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001538}
1539
Avi Kivityc5ea7662007-02-20 18:41:05 +02001540/*
1541 * Creates some virtual cpus. Good luck creating more than one.
1542 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001543static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001544{
1545 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001546 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001547
Gleb Natapov73880c82009-06-09 15:56:28 +03001548 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001549 if (IS_ERR(vcpu))
1550 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001551
Avi Kivity15ad7142007-07-11 18:17:21 +03001552 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1553
Avi Kivity26e52152007-11-20 15:30:24 +02001554 r = kvm_arch_vcpu_setup(vcpu);
1555 if (r)
Glauber Costa7d8fece2008-09-17 23:16:59 -03001556 return r;
Avi Kivity26e52152007-11-20 15:30:24 +02001557
Shaohua Li11ec2802007-07-23 14:51:37 +08001558 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001559 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1560 r = -EINVAL;
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001561 goto vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001562 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001563
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001564 kvm_for_each_vcpu(r, v, kvm)
1565 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001566 r = -EEXIST;
1567 goto vcpu_destroy;
1568 }
1569
1570 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001571
1572 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001573 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001574 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001575 if (r < 0) {
1576 kvm_put_kvm(kvm);
1577 goto vcpu_destroy;
1578 }
1579
1580 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1581 smp_wmb();
1582 atomic_inc(&kvm->online_vcpus);
1583
1584#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1585 if (kvm->bsp_vcpu_id == id)
1586 kvm->bsp_vcpu = vcpu;
1587#endif
1588 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001589 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001590
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08001591vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001592 mutex_unlock(&kvm->lock);
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001593 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001594 return r;
1595}
1596
Avi Kivity1961d272007-03-05 19:46:05 +02001597static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1598{
1599 if (sigset) {
1600 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1601 vcpu->sigset_active = 1;
1602 vcpu->sigset = *sigset;
1603 } else
1604 vcpu->sigset_active = 0;
1605 return 0;
1606}
1607
Avi Kivitybccf2152007-02-21 18:04:26 +02001608static long kvm_vcpu_ioctl(struct file *filp,
1609 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001610{
Avi Kivitybccf2152007-02-21 18:04:26 +02001611 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f366982007-02-09 16:38:35 +00001612 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001613 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001614 struct kvm_fpu *fpu = NULL;
1615 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001616
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001617 if (vcpu->kvm->mm != current->mm)
1618 return -EIO;
Avi Kivity2122ff52010-05-13 11:25:04 +03001619
1620#if defined(CONFIG_S390) || defined(CONFIG_PPC)
1621 /*
1622 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
1623 * so vcpu_load() would break it.
1624 */
1625 if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT)
1626 return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
1627#endif
1628
1629
1630 vcpu_load(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001631 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001632 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001633 r = -EINVAL;
1634 if (arg)
1635 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001636 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Gleb Natapov64be5002010-10-24 16:49:08 +02001637 trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001638 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001639 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001640 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001641
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001642 r = -ENOMEM;
1643 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1644 if (!kvm_regs)
1645 goto out;
1646 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001647 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001648 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001649 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001650 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1651 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001652 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001653out_free1:
1654 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001655 break;
1656 }
1657 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001658 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001659
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001660 r = -ENOMEM;
1661 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1662 if (!kvm_regs)
1663 goto out;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001664 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001665 if (copy_from_user(kvm_regs, argp, sizeof(struct kvm_regs)))
1666 goto out_free2;
1667 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001668 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001669 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001670 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001671out_free2:
1672 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001673 break;
1674 }
1675 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001676 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1677 r = -ENOMEM;
1678 if (!kvm_sregs)
1679 goto out;
1680 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001681 if (r)
1682 goto out;
1683 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001684 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001685 goto out;
1686 r = 0;
1687 break;
1688 }
1689 case KVM_SET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001690 kvm_sregs = kmalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1691 r = -ENOMEM;
1692 if (!kvm_sregs)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001693 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001694 r = -EFAULT;
1695 if (copy_from_user(kvm_sregs, argp, sizeof(struct kvm_sregs)))
1696 goto out;
1697 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001698 if (r)
1699 goto out;
1700 r = 0;
1701 break;
1702 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001703 case KVM_GET_MP_STATE: {
1704 struct kvm_mp_state mp_state;
1705
1706 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1707 if (r)
1708 goto out;
1709 r = -EFAULT;
1710 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1711 goto out;
1712 r = 0;
1713 break;
1714 }
1715 case KVM_SET_MP_STATE: {
1716 struct kvm_mp_state mp_state;
1717
1718 r = -EFAULT;
1719 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1720 goto out;
1721 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1722 if (r)
1723 goto out;
1724 r = 0;
1725 break;
1726 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001727 case KVM_TRANSLATE: {
1728 struct kvm_translation tr;
1729
1730 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001731 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001732 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001733 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001734 if (r)
1735 goto out;
1736 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001737 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001738 goto out;
1739 r = 0;
1740 break;
1741 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001742 case KVM_SET_GUEST_DEBUG: {
1743 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001744
1745 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001746 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001747 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001748 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001749 if (r)
1750 goto out;
1751 r = 0;
1752 break;
1753 }
Avi Kivity1961d272007-03-05 19:46:05 +02001754 case KVM_SET_SIGNAL_MASK: {
1755 struct kvm_signal_mask __user *sigmask_arg = argp;
1756 struct kvm_signal_mask kvm_sigmask;
1757 sigset_t sigset, *p;
1758
1759 p = NULL;
1760 if (argp) {
1761 r = -EFAULT;
1762 if (copy_from_user(&kvm_sigmask, argp,
1763 sizeof kvm_sigmask))
1764 goto out;
1765 r = -EINVAL;
1766 if (kvm_sigmask.len != sizeof sigset)
1767 goto out;
1768 r = -EFAULT;
1769 if (copy_from_user(&sigset, sigmask_arg->sigset,
1770 sizeof sigset))
1771 goto out;
1772 p = &sigset;
1773 }
Andi Kleen376d41f2010-06-10 13:10:47 +02001774 r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
Avi Kivity1961d272007-03-05 19:46:05 +02001775 break;
1776 }
Avi Kivityb8836732007-04-01 16:34:31 +03001777 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001778 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1779 r = -ENOMEM;
1780 if (!fpu)
1781 goto out;
1782 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001783 if (r)
1784 goto out;
1785 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001786 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001787 goto out;
1788 r = 0;
1789 break;
1790 }
1791 case KVM_SET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001792 fpu = kmalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1793 r = -ENOMEM;
1794 if (!fpu)
Avi Kivityb8836732007-04-01 16:34:31 +03001795 goto out;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001796 r = -EFAULT;
1797 if (copy_from_user(fpu, argp, sizeof(struct kvm_fpu)))
1798 goto out;
1799 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001800 if (r)
1801 goto out;
1802 r = 0;
1803 break;
1804 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001805 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001806 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001807 }
1808out:
Avi Kivity2122ff52010-05-13 11:25:04 +03001809 vcpu_put(vcpu);
Dave Hansenfa3795a2008-08-11 10:01:46 -07001810 kfree(fpu);
1811 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001812 return r;
1813}
1814
1815static long kvm_vm_ioctl(struct file *filp,
1816 unsigned int ioctl, unsigned long arg)
1817{
1818 struct kvm *kvm = filp->private_data;
1819 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001820 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001821
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001822 if (kvm->mm != current->mm)
1823 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001824 switch (ioctl) {
1825 case KVM_CREATE_VCPU:
1826 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1827 if (r < 0)
1828 goto out;
1829 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001830 case KVM_SET_USER_MEMORY_REGION: {
1831 struct kvm_userspace_memory_region kvm_userspace_mem;
1832
1833 r = -EFAULT;
1834 if (copy_from_user(&kvm_userspace_mem, argp,
1835 sizeof kvm_userspace_mem))
1836 goto out;
1837
1838 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001839 if (r)
1840 goto out;
1841 break;
1842 }
1843 case KVM_GET_DIRTY_LOG: {
1844 struct kvm_dirty_log log;
1845
1846 r = -EFAULT;
Al Viro2f366982007-02-09 16:38:35 +00001847 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001848 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02001849 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001850 if (r)
1851 goto out;
1852 break;
1853 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02001854#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1855 case KVM_REGISTER_COALESCED_MMIO: {
1856 struct kvm_coalesced_mmio_zone zone;
1857 r = -EFAULT;
1858 if (copy_from_user(&zone, argp, sizeof zone))
1859 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02001860 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
1861 if (r)
1862 goto out;
1863 r = 0;
1864 break;
1865 }
1866 case KVM_UNREGISTER_COALESCED_MMIO: {
1867 struct kvm_coalesced_mmio_zone zone;
1868 r = -EFAULT;
1869 if (copy_from_user(&zone, argp, sizeof zone))
1870 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02001871 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
1872 if (r)
1873 goto out;
1874 r = 0;
1875 break;
1876 }
1877#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04001878 case KVM_IRQFD: {
1879 struct kvm_irqfd data;
1880
1881 r = -EFAULT;
1882 if (copy_from_user(&data, argp, sizeof data))
1883 goto out;
1884 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
1885 break;
1886 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04001887 case KVM_IOEVENTFD: {
1888 struct kvm_ioeventfd data;
1889
1890 r = -EFAULT;
1891 if (copy_from_user(&data, argp, sizeof data))
1892 goto out;
1893 r = kvm_ioeventfd(kvm, &data);
1894 break;
1895 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001896#ifdef CONFIG_KVM_APIC_ARCHITECTURE
1897 case KVM_SET_BOOT_CPU_ID:
1898 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03001899 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001900 if (atomic_read(&kvm->online_vcpus) != 0)
1901 r = -EBUSY;
1902 else
1903 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03001904 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03001905 break;
1906#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001907 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01001908 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03001909 if (r == -ENOTTY)
1910 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02001911 }
1912out:
1913 return r;
1914}
1915
Arnd Bergmann6ff58942009-10-22 14:19:27 +02001916#ifdef CONFIG_COMPAT
1917struct compat_kvm_dirty_log {
1918 __u32 slot;
1919 __u32 padding1;
1920 union {
1921 compat_uptr_t dirty_bitmap; /* one bit per page */
1922 __u64 padding2;
1923 };
1924};
1925
1926static long kvm_vm_compat_ioctl(struct file *filp,
1927 unsigned int ioctl, unsigned long arg)
1928{
1929 struct kvm *kvm = filp->private_data;
1930 int r;
1931
1932 if (kvm->mm != current->mm)
1933 return -EIO;
1934 switch (ioctl) {
1935 case KVM_GET_DIRTY_LOG: {
1936 struct compat_kvm_dirty_log compat_log;
1937 struct kvm_dirty_log log;
1938
1939 r = -EFAULT;
1940 if (copy_from_user(&compat_log, (void __user *)arg,
1941 sizeof(compat_log)))
1942 goto out;
1943 log.slot = compat_log.slot;
1944 log.padding1 = compat_log.padding1;
1945 log.padding2 = compat_log.padding2;
1946 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
1947
1948 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
1949 if (r)
1950 goto out;
1951 break;
1952 }
1953 default:
1954 r = kvm_vm_ioctl(filp, ioctl, arg);
1955 }
1956
1957out:
1958 return r;
1959}
1960#endif
1961
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001962static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02001963{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001964 struct page *page[1];
1965 unsigned long addr;
1966 int npages;
1967 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02001968 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02001969
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001970 addr = gfn_to_hva(kvm, gfn);
1971 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001972 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001973
1974 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
1975 NULL);
1976 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001977 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03001978
1979 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001980 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02001981}
1982
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001983static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001984 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02001985};
1986
1987static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
1988{
1989 vma->vm_ops = &kvm_vm_vm_ops;
1990 return 0;
1991}
1992
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001993static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02001994 .release = kvm_vm_release,
1995 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02001996#ifdef CONFIG_COMPAT
1997 .compat_ioctl = kvm_vm_compat_ioctl,
1998#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02001999 .mmap = kvm_vm_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002000 .llseek = noop_llseek,
Avi Kivityf17abe92007-02-21 19:28:04 +02002001};
2002
2003static int kvm_dev_ioctl_create_vm(void)
2004{
Heiko Carstensaac87632010-10-27 17:22:10 +02002005 int r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002006 struct kvm *kvm;
2007
Avi Kivityf17abe92007-02-21 19:28:04 +02002008 kvm = kvm_create_vm();
Avi Kivityd6d28162007-06-28 08:38:16 -04002009 if (IS_ERR(kvm))
2010 return PTR_ERR(kvm);
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09002011#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2012 r = kvm_coalesced_mmio_init(kvm);
2013 if (r < 0) {
2014 kvm_put_kvm(kvm);
2015 return r;
2016 }
2017#endif
Heiko Carstensaac87632010-10-27 17:22:10 +02002018 r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
2019 if (r < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002020 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002021
Heiko Carstensaac87632010-10-27 17:22:10 +02002022 return r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002023}
2024
Avi Kivity1a811b62008-12-08 18:25:27 +02002025static long kvm_dev_ioctl_check_extension_generic(long arg)
2026{
2027 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002028 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002029 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002030 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002031#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2032 case KVM_CAP_SET_BOOT_CPU_ID:
2033#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02002034 case KVM_CAP_INTERNAL_ERROR_DATA:
Avi Kivity1a811b62008-12-08 18:25:27 +02002035 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02002036#ifdef CONFIG_HAVE_KVM_IRQCHIP
2037 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002038 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002039#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002040 default:
2041 break;
2042 }
2043 return kvm_dev_ioctl_check_extension(arg);
2044}
2045
Avi Kivityf17abe92007-02-21 19:28:04 +02002046static long kvm_dev_ioctl(struct file *filp,
2047 unsigned int ioctl, unsigned long arg)
2048{
Avi Kivity07c45a32007-03-07 13:05:38 +02002049 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002050
2051 switch (ioctl) {
2052 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002053 r = -EINVAL;
2054 if (arg)
2055 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002056 r = KVM_API_VERSION;
2057 break;
2058 case KVM_CREATE_VM:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002059 r = -EINVAL;
2060 if (arg)
2061 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002062 r = kvm_dev_ioctl_create_vm();
2063 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002064 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002065 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002066 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002067 case KVM_GET_VCPU_MMAP_SIZE:
2068 r = -EINVAL;
2069 if (arg)
2070 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002071 r = PAGE_SIZE; /* struct kvm_run */
2072#ifdef CONFIG_X86
2073 r += PAGE_SIZE; /* pio data page */
2074#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002075#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2076 r += PAGE_SIZE; /* coalesced mmio ring page */
2077#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002078 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002079 case KVM_TRACE_ENABLE:
2080 case KVM_TRACE_PAUSE:
2081 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002082 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002083 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002084 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002085 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002086 }
2087out:
2088 return r;
2089}
2090
Avi Kivity6aa8b732006-12-10 02:21:36 -08002091static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002092 .unlocked_ioctl = kvm_dev_ioctl,
2093 .compat_ioctl = kvm_dev_ioctl,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002094 .llseek = noop_llseek,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002095};
2096
2097static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002098 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002099 "kvm",
2100 &kvm_chardev_ops,
2101};
2102
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002103static void hardware_enable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002104{
2105 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02002106 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03002107
Rusty Russell7f59f492008-12-07 21:25:45 +10302108 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002109 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02002110
Rusty Russell7f59f492008-12-07 21:25:45 +10302111 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02002112
2113 r = kvm_arch_hardware_enable(NULL);
2114
2115 if (r) {
2116 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
2117 atomic_inc(&hardware_enable_failed);
2118 printk(KERN_INFO "kvm: enabling virtualization on "
2119 "CPU%d failed\n", cpu);
2120 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03002121}
2122
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002123static void hardware_enable(void *junk)
2124{
2125 spin_lock(&kvm_lock);
2126 hardware_enable_nolock(junk);
2127 spin_unlock(&kvm_lock);
2128}
2129
2130static void hardware_disable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002131{
2132 int cpu = raw_smp_processor_id();
2133
Rusty Russell7f59f492008-12-07 21:25:45 +10302134 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002135 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302136 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002137 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002138}
2139
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002140static void hardware_disable(void *junk)
2141{
2142 spin_lock(&kvm_lock);
2143 hardware_disable_nolock(junk);
2144 spin_unlock(&kvm_lock);
2145}
2146
Alexander Graf10474ae2009-09-15 11:37:46 +02002147static void hardware_disable_all_nolock(void)
2148{
2149 BUG_ON(!kvm_usage_count);
2150
2151 kvm_usage_count--;
2152 if (!kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002153 on_each_cpu(hardware_disable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002154}
2155
2156static void hardware_disable_all(void)
2157{
2158 spin_lock(&kvm_lock);
2159 hardware_disable_all_nolock();
2160 spin_unlock(&kvm_lock);
2161}
2162
2163static int hardware_enable_all(void)
2164{
2165 int r = 0;
2166
2167 spin_lock(&kvm_lock);
2168
2169 kvm_usage_count++;
2170 if (kvm_usage_count == 1) {
2171 atomic_set(&hardware_enable_failed, 0);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002172 on_each_cpu(hardware_enable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002173
2174 if (atomic_read(&hardware_enable_failed)) {
2175 hardware_disable_all_nolock();
2176 r = -EBUSY;
2177 }
2178 }
2179
2180 spin_unlock(&kvm_lock);
2181
2182 return r;
2183}
2184
Avi Kivity774c47f2007-02-12 00:54:47 -08002185static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2186 void *v)
2187{
2188 int cpu = (long)v;
2189
Alexander Graf10474ae2009-09-15 11:37:46 +02002190 if (!kvm_usage_count)
2191 return NOTIFY_OK;
2192
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002193 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002194 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002195 case CPU_DYING:
Avi Kivity6ec8a852007-08-19 15:57:26 +03002196 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2197 cpu);
2198 hardware_disable(NULL);
2199 break;
Zachary Amsdenda908f22010-08-19 22:07:27 -10002200 case CPU_STARTING:
Jeremy Katz43934a32007-02-19 14:37:46 +02002201 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2202 cpu);
Zachary Amsdenda908f22010-08-19 22:07:27 -10002203 hardware_enable(NULL);
Avi Kivity774c47f2007-02-12 00:54:47 -08002204 break;
2205 }
2206 return NOTIFY_OK;
2207}
2208
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002209
Avi Kivityb7c41452010-12-02 17:52:50 +02002210asmlinkage void kvm_spurious_fault(void)
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002211{
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002212 /* Fault while not rebooting. We want the trace. */
2213 BUG();
2214}
Avi Kivityb7c41452010-12-02 17:52:50 +02002215EXPORT_SYMBOL_GPL(kvm_spurious_fault);
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002216
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002217static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002218 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002219{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002220 /*
2221 * Some (well, at least mine) BIOSes hang on reboot if
2222 * in vmx root mode.
2223 *
2224 * And Intel TXT required VMX off for all cpu when system shutdown.
2225 */
2226 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2227 kvm_rebooting = true;
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002228 on_each_cpu(hardware_disable_nolock, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002229 return NOTIFY_OK;
2230}
2231
2232static struct notifier_block kvm_reboot_notifier = {
2233 .notifier_call = kvm_reboot,
2234 .priority = 0,
2235};
2236
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002237static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002238{
2239 int i;
2240
2241 for (i = 0; i < bus->dev_count; i++) {
2242 struct kvm_io_device *pos = bus->devs[i];
2243
2244 kvm_iodevice_destructor(pos);
2245 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002246 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002247}
2248
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002249/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002250int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002251 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002252{
2253 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002254 struct kvm_io_bus *bus;
2255
2256 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002257 for (i = 0; i < bus->dev_count; i++)
2258 if (!kvm_iodevice_write(bus->devs[i], addr, len, val))
2259 return 0;
2260 return -EOPNOTSUPP;
2261}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002262
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002263/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002264int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2265 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002266{
2267 int i;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002268 struct kvm_io_bus *bus;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002269
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002270 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002271 for (i = 0; i < bus->dev_count; i++)
2272 if (!kvm_iodevice_read(bus->devs[i], addr, len, val))
2273 return 0;
2274 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002275}
2276
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002277/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002278int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2279 struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002280{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002281 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002282
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002283 bus = kvm->buses[bus_idx];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002284 if (bus->dev_count > NR_IOBUS_DEVS-1)
2285 return -ENOSPC;
2286
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002287 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2288 if (!new_bus)
2289 return -ENOMEM;
2290 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2291 new_bus->devs[new_bus->dev_count++] = dev;
2292 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2293 synchronize_srcu_expedited(&kvm->srcu);
2294 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002295
2296 return 0;
2297}
2298
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002299/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002300int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2301 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002302{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002303 int i, r;
2304 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002305
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002306 new_bus = kzalloc(sizeof(struct kvm_io_bus), GFP_KERNEL);
2307 if (!new_bus)
2308 return -ENOMEM;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002309
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002310 bus = kvm->buses[bus_idx];
2311 memcpy(new_bus, bus, sizeof(struct kvm_io_bus));
2312
2313 r = -ENOENT;
2314 for (i = 0; i < new_bus->dev_count; i++)
2315 if (new_bus->devs[i] == dev) {
2316 r = 0;
2317 new_bus->devs[i] = new_bus->devs[--new_bus->dev_count];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002318 break;
2319 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002320
2321 if (r) {
2322 kfree(new_bus);
2323 return r;
2324 }
2325
2326 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2327 synchronize_srcu_expedited(&kvm->srcu);
2328 kfree(bus);
2329 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002330}
2331
Avi Kivity774c47f2007-02-12 00:54:47 -08002332static struct notifier_block kvm_cpu_notifier = {
2333 .notifier_call = kvm_cpu_hotplug,
Avi Kivity774c47f2007-02-12 00:54:47 -08002334};
2335
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002336static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002337{
2338 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002339 struct kvm *kvm;
2340
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002341 *val = 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002342 spin_lock(&kvm_lock);
2343 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002344 *val += *(u32 *)((void *)kvm + offset);
Avi Kivityba1389b2007-11-18 16:24:12 +02002345 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002346 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002347}
2348
2349DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2350
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002351static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002352{
2353 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002354 struct kvm *kvm;
2355 struct kvm_vcpu *vcpu;
2356 int i;
2357
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002358 *val = 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002359 spin_lock(&kvm_lock);
2360 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002361 kvm_for_each_vcpu(i, vcpu, kvm)
2362 *val += *(u32 *)((void *)vcpu + offset);
2363
Avi Kivity1165f5f2007-04-19 17:27:43 +03002364 spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002365 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002366}
2367
Avi Kivityba1389b2007-11-18 16:24:12 +02002368DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2369
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002370static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002371 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2372 [KVM_STAT_VM] = &vm_stat_fops,
2373};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002374
Zhang Xiantaoa16b0432007-11-16 14:38:21 +08002375static void kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002376{
2377 struct kvm_stats_debugfs_item *p;
2378
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002379 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002380 for (p = debugfs_entries; p->name; ++p)
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002381 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002382 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002383 stat_fops[p->kind]);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002384}
2385
2386static void kvm_exit_debug(void)
2387{
2388 struct kvm_stats_debugfs_item *p;
2389
2390 for (p = debugfs_entries; p->name; ++p)
2391 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002392 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002393}
2394
Avi Kivity59ae6c62007-02-12 00:54:48 -08002395static int kvm_suspend(struct sys_device *dev, pm_message_t state)
2396{
Alexander Graf10474ae2009-09-15 11:37:46 +02002397 if (kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002398 hardware_disable_nolock(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002399 return 0;
2400}
2401
2402static int kvm_resume(struct sys_device *dev)
2403{
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002404 if (kvm_usage_count) {
2405 WARN_ON(spin_is_locked(&kvm_lock));
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002406 hardware_enable_nolock(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002407 }
Avi Kivity59ae6c62007-02-12 00:54:48 -08002408 return 0;
2409}
2410
2411static struct sysdev_class kvm_sysdev_class = {
Kay Sieversaf5ca3f2007-12-20 02:09:39 +01002412 .name = "kvm",
Avi Kivity59ae6c62007-02-12 00:54:48 -08002413 .suspend = kvm_suspend,
2414 .resume = kvm_resume,
2415};
2416
2417static struct sys_device kvm_sysdev = {
2418 .id = 0,
2419 .cls = &kvm_sysdev_class,
2420};
2421
Izik Eiduscea7bb22007-10-17 19:17:48 +02002422struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002423pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002424
Avi Kivity15ad7142007-07-11 18:17:21 +03002425static inline
2426struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2427{
2428 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2429}
2430
2431static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2432{
2433 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2434
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002435 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002436}
2437
2438static void kvm_sched_out(struct preempt_notifier *pn,
2439 struct task_struct *next)
2440{
2441 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2442
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002443 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002444}
2445
Avi Kivity0ee75be2010-04-28 15:39:01 +03002446int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
Rusty Russellc16f8622007-07-30 21:12:19 +10002447 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002448{
2449 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002450 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002451
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002452 r = kvm_arch_init(opaque);
2453 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002454 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002455
2456 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2457
2458 if (bad_page == NULL) {
2459 r = -ENOMEM;
2460 goto out;
2461 }
2462
Anthony Liguori35149e22008-04-02 14:46:56 -05002463 bad_pfn = page_to_pfn(bad_page);
2464
Huang Yingbf998152010-05-31 14:28:19 +08002465 hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2466
2467 if (hwpoison_page == NULL) {
2468 r = -ENOMEM;
2469 goto out_free_0;
2470 }
2471
2472 hwpoison_pfn = page_to_pfn(hwpoison_page);
2473
Gleb Natapovedba23e2010-07-07 20:16:45 +03002474 fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2475
2476 if (fault_page == NULL) {
2477 r = -ENOMEM;
2478 goto out_free_0;
2479 }
2480
2481 fault_pfn = page_to_pfn(fault_page);
2482
Avi Kivity8437a612009-06-06 14:52:35 -07002483 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302484 r = -ENOMEM;
2485 goto out_free_0;
2486 }
2487
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002488 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002489 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302490 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002491
Yang, Sheng002c7f72007-07-31 14:23:01 +03002492 for_each_online_cpu(cpu) {
2493 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002494 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002495 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002496 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002497 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002498 }
2499
Avi Kivity774c47f2007-02-12 00:54:47 -08002500 r = register_cpu_notifier(&kvm_cpu_notifier);
2501 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002502 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002503 register_reboot_notifier(&kvm_reboot_notifier);
2504
Avi Kivity59ae6c62007-02-12 00:54:48 -08002505 r = sysdev_class_register(&kvm_sysdev_class);
2506 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002507 goto out_free_3;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002508
2509 r = sysdev_register(&kvm_sysdev);
2510 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002511 goto out_free_4;
Avi Kivity59ae6c62007-02-12 00:54:48 -08002512
Rusty Russellc16f8622007-07-30 21:12:19 +10002513 /* A kmem cache lets us meet the alignment requirements of fx_save. */
Avi Kivity0ee75be2010-04-28 15:39:01 +03002514 if (!vcpu_align)
2515 vcpu_align = __alignof__(struct kvm_vcpu);
2516 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
Joe Perches56919c52007-11-12 20:06:51 -08002517 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002518 if (!kvm_vcpu_cache) {
2519 r = -ENOMEM;
Zhang Xiantaod23087842007-11-29 15:35:39 +08002520 goto out_free_5;
Rusty Russellc16f8622007-07-30 21:12:19 +10002521 }
2522
Gleb Natapovaf585b92010-10-14 11:22:46 +02002523 r = kvm_async_pf_init();
2524 if (r)
2525 goto out_free;
2526
Avi Kivity6aa8b732006-12-10 02:21:36 -08002527 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002528 kvm_vm_fops.owner = module;
2529 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002530
2531 r = misc_register(&kvm_dev);
2532 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002533 printk(KERN_ERR "kvm: misc device register failed\n");
Gleb Natapovaf585b92010-10-14 11:22:46 +02002534 goto out_unreg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002535 }
2536
Avi Kivity15ad7142007-07-11 18:17:21 +03002537 kvm_preempt_ops.sched_in = kvm_sched_in;
2538 kvm_preempt_ops.sched_out = kvm_sched_out;
2539
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002540 kvm_init_debug();
2541
Avi Kivityc7addb92007-09-16 18:58:32 +02002542 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002543
Gleb Natapovaf585b92010-10-14 11:22:46 +02002544out_unreg:
2545 kvm_async_pf_deinit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002546out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002547 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002548out_free_5:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002549 sysdev_unregister(&kvm_sysdev);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002550out_free_4:
Avi Kivity59ae6c62007-02-12 00:54:48 -08002551 sysdev_class_unregister(&kvm_sysdev_class);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002552out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002553 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002554 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002555out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002556out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002557 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302558out_free_0a:
2559 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002560out_free_0:
Gleb Natapovedba23e2010-07-07 20:16:45 +03002561 if (fault_page)
2562 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002563 if (hwpoison_page)
2564 __free_page(hwpoison_page);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002565 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002566out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002567 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002568out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002569 return r;
2570}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002571EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002572
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002573void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002574{
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002575 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002576 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002577 kmem_cache_destroy(kvm_vcpu_cache);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002578 kvm_async_pf_deinit();
Avi Kivity59ae6c62007-02-12 00:54:48 -08002579 sysdev_unregister(&kvm_sysdev);
2580 sysdev_class_unregister(&kvm_sysdev_class);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002581 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002582 unregister_cpu_notifier(&kvm_cpu_notifier);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002583 on_each_cpu(hardware_disable_nolock, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002584 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002585 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302586 free_cpumask_var(cpus_hardware_enabled);
Huang Yingbf998152010-05-31 14:28:19 +08002587 __free_page(hwpoison_page);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002588 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002589}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002590EXPORT_SYMBOL_GPL(kvm_exit);