blob: 9739b533ca2e6954f75c98fc1bb307e641aedf94 [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>
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +010033#include <linux/syscore_ops.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>
Sasha Levin743eeb02011-07-27 16:00:48 +030050#include <linux/sort.h>
51#include <linux/bsearch.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080052
Avi Kivitye4956062007-06-28 14:15:57 -040053#include <asm/processor.h>
Avi Kivitye4956062007-06-28 14:15:57 -040054#include <asm/io.h>
55#include <asm/uaccess.h>
Izik Eidus3e021bf2007-11-19 11:16:57 +020056#include <asm/pgtable.h>
Avi Kivity6aa8b732006-12-10 02:21:36 -080057
Laurent Vivier5f94c172008-05-30 16:05:54 +020058#include "coalesced_mmio.h"
Gleb Natapovaf585b92010-10-14 11:22:46 +020059#include "async_pf.h"
Laurent Vivier5f94c172008-05-30 16:05:54 +020060
Marcelo Tosatti229456f2009-06-17 09:22:14 -030061#define CREATE_TRACE_POINTS
62#include <trace/events/kvm.h>
63
Avi Kivity6aa8b732006-12-10 02:21:36 -080064MODULE_AUTHOR("Qumranet");
65MODULE_LICENSE("GPL");
66
Marcelo Tosattifa40a822009-06-04 15:08:24 -030067/*
68 * Ordering of locks:
69 *
Sheng Yangfae3a352009-12-15 10:28:07 +080070 * kvm->lock --> kvm->slots_lock --> kvm->irq_lock
Marcelo Tosattifa40a822009-06-04 15:08:24 -030071 */
72
Jan Kiszkae935b832011-02-08 12:55:33 +010073DEFINE_RAW_SPINLOCK(kvm_lock);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +080074LIST_HEAD(vm_list);
Avi Kivity133de902007-02-12 00:54:44 -080075
Rusty Russell7f59f492008-12-07 21:25:45 +103076static cpumask_var_t cpus_hardware_enabled;
Alexander Graf10474ae2009-09-15 11:37:46 +020077static int kvm_usage_count = 0;
78static atomic_t hardware_enable_failed;
Avi Kivity1b6c0162007-05-24 13:03:52 +030079
Rusty Russellc16f8622007-07-30 21:12:19 +100080struct kmem_cache *kvm_vcpu_cache;
81EXPORT_SYMBOL_GPL(kvm_vcpu_cache);
Avi Kivity1165f5f2007-04-19 17:27:43 +030082
Avi Kivity15ad7142007-07-11 18:17:21 +030083static __read_mostly struct preempt_ops kvm_preempt_ops;
84
Hollis Blanchard76f7c872008-04-15 16:05:42 -050085struct dentry *kvm_debugfs_dir;
Avi Kivity6aa8b732006-12-10 02:21:36 -080086
Avi Kivitybccf2152007-02-21 18:04:26 +020087static long kvm_vcpu_ioctl(struct file *file, unsigned int ioctl,
88 unsigned long arg);
Alexander Graf1dda6062011-06-08 02:45:37 +020089#ifdef CONFIG_COMPAT
90static long kvm_vcpu_compat_ioctl(struct file *file, unsigned int ioctl,
91 unsigned long arg);
92#endif
Alexander Graf10474ae2009-09-15 11:37:46 +020093static int hardware_enable_all(void);
94static void hardware_disable_all(void);
Avi Kivitybccf2152007-02-21 18:04:26 +020095
Marcelo Tosattie93f8a02009-12-23 14:35:24 -020096static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
97
Avi Kivityb7c41452010-12-02 17:52:50 +020098bool kvm_rebooting;
99EXPORT_SYMBOL_GPL(kvm_rebooting);
Avi Kivity4ecac3f2008-05-13 13:23:38 +0300100
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300101static bool largepages_enabled = true;
102
Gleb Natapovfa7bff82010-07-07 20:16:44 +0300103static struct page *hwpoison_page;
104static pfn_t hwpoison_pfn;
Huang Yingbf998152010-05-31 14:28:19 +0800105
Xiao Guangrongfce92dc2011-07-12 03:28:54 +0800106struct page *fault_page;
107pfn_t fault_pfn;
Gleb Natapovedba23e2010-07-07 20:16:45 +0300108
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +0800109inline int kvm_is_mmio_pfn(pfn_t pfn)
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300110{
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100111 if (pfn_valid(pfn)) {
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800112 int reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800113 struct page *tail = pfn_to_page(pfn);
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800114 struct page *head = compound_trans_head(tail);
115 reserved = PageReserved(head);
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800116 if (head != tail) {
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800117 /*
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800118 * "head" is not a dangling pointer
119 * (compound_trans_head takes care of that)
120 * but the hugepage may have been splitted
121 * from under us (and we may not hold a
122 * reference count on the head page so it can
123 * be reused before we run PageReferenced), so
124 * we've to check PageTail before returning
125 * what we just read.
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800126 */
Andrea Arcangeli22e5c472011-01-13 15:47:20 -0800127 smp_rmb();
128 if (PageTail(tail))
129 return reserved;
Andrea Arcangeli936a5fe2011-01-13 15:46:48 -0800130 }
131 return PageReserved(tail);
Joerg Roedelfc5659c2009-02-18 14:08:58 +0100132 }
Ben-Ami Yassourcbff90a2008-07-28 19:26:24 +0300133
134 return true;
135}
136
Avi Kivity6aa8b732006-12-10 02:21:36 -0800137/*
138 * Switches to specified vcpu, until a matching vcpu_put()
139 */
Carsten Otte313a3dc2007-10-11 19:16:52 +0200140void vcpu_load(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800141{
Avi Kivity15ad7142007-07-11 18:17:21 +0300142 int cpu;
143
Avi Kivitybccf2152007-02-21 18:04:26 +0200144 mutex_lock(&vcpu->mutex);
Rik van Riel34bb10b2011-02-01 09:52:41 -0500145 if (unlikely(vcpu->pid != current->pids[PIDTYPE_PID].pid)) {
146 /* The thread running this VCPU changed. */
147 struct pid *oldpid = vcpu->pid;
148 struct pid *newpid = get_task_pid(current, PIDTYPE_PID);
149 rcu_assign_pointer(vcpu->pid, newpid);
150 synchronize_rcu();
151 put_pid(oldpid);
152 }
Avi Kivity15ad7142007-07-11 18:17:21 +0300153 cpu = get_cpu();
154 preempt_notifier_register(&vcpu->preempt_notifier);
Carsten Otte313a3dc2007-10-11 19:16:52 +0200155 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300156 put_cpu();
Avi Kivitybccf2152007-02-21 18:04:26 +0200157}
158
Carsten Otte313a3dc2007-10-11 19:16:52 +0200159void vcpu_put(struct kvm_vcpu *vcpu)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800160{
Avi Kivity15ad7142007-07-11 18:17:21 +0300161 preempt_disable();
Carsten Otte313a3dc2007-10-11 19:16:52 +0200162 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +0300163 preempt_notifier_unregister(&vcpu->preempt_notifier);
164 preempt_enable();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800165 mutex_unlock(&vcpu->mutex);
166}
167
Avi Kivityd9e368d2007-06-07 19:18:30 +0300168static void ack_flush(void *_completed)
169{
Avi Kivityd9e368d2007-06-07 19:18:30 +0300170}
171
Rusty Russell49846892008-12-08 20:26:24 +1030172static bool make_all_cpus_request(struct kvm *kvm, unsigned int req)
Avi Kivityd9e368d2007-06-07 19:18:30 +0300173{
Avi Kivity597a5f52008-07-20 14:24:22 +0300174 int i, cpu, me;
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030175 cpumask_var_t cpus;
176 bool called = true;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300177 struct kvm_vcpu *vcpu;
Avi Kivityd9e368d2007-06-07 19:18:30 +0300178
Li Zefan79f55992009-06-15 14:58:26 +0800179 zalloc_cpumask_var(&cpus, GFP_ATOMIC);
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030180
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800181 me = get_cpu();
Gleb Natapov988a2ca2009-06-09 15:56:29 +0300182 kvm_for_each_vcpu(i, vcpu, kvm) {
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800183 kvm_make_request(req, vcpu);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300184 cpu = vcpu->cpu;
Xiao Guangrong6b7e2d02011-01-12 15:40:31 +0800185
186 /* Set ->requests bit before we read ->mode */
187 smp_mb();
188
189 if (cpus != NULL && cpu != -1 && cpu != me &&
190 kvm_vcpu_exiting_guest_mode(vcpu) != OUTSIDE_GUEST_MODE)
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030191 cpumask_set_cpu(cpu, cpus);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300192 }
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030193 if (unlikely(cpus == NULL))
194 smp_call_function_many(cpu_online_mask, ack_flush, NULL, 1);
195 else if (!cpumask_empty(cpus))
196 smp_call_function_many(cpus, ack_flush, NULL, 1);
197 else
198 called = false;
Xiao Guangrong3cba4132011-01-12 15:41:22 +0800199 put_cpu();
Rusty Russell6ef7a1b2008-12-08 20:28:04 +1030200 free_cpumask_var(cpus);
Rusty Russell49846892008-12-08 20:26:24 +1030201 return called;
202}
203
204void kvm_flush_remote_tlbs(struct kvm *kvm)
205{
Alex Shibec87d62012-03-04 13:13:38 +0800206 long dirty_count = kvm->tlbs_dirty;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800207
208 smp_mb();
Rusty Russell49846892008-12-08 20:26:24 +1030209 if (make_all_cpus_request(kvm, KVM_REQ_TLB_FLUSH))
210 ++kvm->stat.remote_tlb_flush;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800211 cmpxchg(&kvm->tlbs_dirty, dirty_count, 0);
Avi Kivityd9e368d2007-06-07 19:18:30 +0300212}
213
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500214void kvm_reload_remote_mmus(struct kvm *kvm)
215{
Rusty Russell49846892008-12-08 20:26:24 +1030216 make_all_cpus_request(kvm, KVM_REQ_MMU_RELOAD);
Marcelo Tosatti2e53d632008-02-20 14:47:24 -0500217}
218
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000219int kvm_vcpu_init(struct kvm_vcpu *vcpu, struct kvm *kvm, unsigned id)
220{
221 struct page *page;
222 int r;
223
224 mutex_init(&vcpu->mutex);
225 vcpu->cpu = -1;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000226 vcpu->kvm = kvm;
227 vcpu->vcpu_id = id;
Rik van Riel34bb10b2011-02-01 09:52:41 -0500228 vcpu->pid = NULL;
Eddie Dongb6958ce2007-07-18 12:15:21 +0300229 init_waitqueue_head(&vcpu->wq);
Gleb Natapovaf585b92010-10-14 11:22:46 +0200230 kvm_async_pf_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000231
232 page = alloc_page(GFP_KERNEL | __GFP_ZERO);
233 if (!page) {
234 r = -ENOMEM;
235 goto fail;
236 }
237 vcpu->run = page_address(page);
238
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800239 r = kvm_arch_vcpu_init(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000240 if (r < 0)
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800241 goto fail_free_run;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000242 return 0;
243
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000244fail_free_run:
245 free_page((unsigned long)vcpu->run);
246fail:
Rusty Russell76fafa52007-10-08 10:50:48 +1000247 return r;
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000248}
249EXPORT_SYMBOL_GPL(kvm_vcpu_init);
250
251void kvm_vcpu_uninit(struct kvm_vcpu *vcpu)
252{
Rik van Riel34bb10b2011-02-01 09:52:41 -0500253 put_pid(vcpu->pid);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +0800254 kvm_arch_vcpu_uninit(vcpu);
Rusty Russellfb3f0f52007-07-27 17:16:56 +1000255 free_page((unsigned long)vcpu->run);
256}
257EXPORT_SYMBOL_GPL(kvm_vcpu_uninit);
258
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200259#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
260static inline struct kvm *mmu_notifier_to_kvm(struct mmu_notifier *mn)
261{
262 return container_of(mn, struct kvm, mmu_notifier);
263}
264
265static void kvm_mmu_notifier_invalidate_page(struct mmu_notifier *mn,
266 struct mm_struct *mm,
267 unsigned long address)
268{
269 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200270 int need_tlb_flush, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200271
272 /*
273 * When ->invalidate_page runs, the linux pte has been zapped
274 * already but the page is still allocated until
275 * ->invalidate_page returns. So if we increase the sequence
276 * here the kvm page fault will notice if the spte can't be
277 * established because the page is going to be freed. If
278 * instead the kvm page fault establishes the spte before
279 * ->invalidate_page runs, kvm_unmap_hva will release it
280 * before returning.
281 *
282 * The sequence increase only need to be seen at spin_unlock
283 * time, and not at spin_lock time.
284 *
285 * Increasing the sequence after the spin_unlock would be
286 * unsafe because the kvm page fault could then establish the
287 * pte after kvm_unmap_hva returned, without noticing the page
288 * is going to be freed.
289 */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200290 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200291 spin_lock(&kvm->mmu_lock);
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900292
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200293 kvm->mmu_notifier_seq++;
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800294 need_tlb_flush = kvm_unmap_hva(kvm, address) | kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200295 /* we've to flush the tlb before the pages can be freed */
296 if (need_tlb_flush)
297 kvm_flush_remote_tlbs(kvm);
298
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900299 spin_unlock(&kvm->mmu_lock);
300 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200301}
302
Izik Eidus3da0dd42009-09-23 21:47:18 +0300303static void kvm_mmu_notifier_change_pte(struct mmu_notifier *mn,
304 struct mm_struct *mm,
305 unsigned long address,
306 pte_t pte)
307{
308 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200309 int idx;
Izik Eidus3da0dd42009-09-23 21:47:18 +0300310
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200311 idx = srcu_read_lock(&kvm->srcu);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300312 spin_lock(&kvm->mmu_lock);
313 kvm->mmu_notifier_seq++;
314 kvm_set_spte_hva(kvm, address, pte);
315 spin_unlock(&kvm->mmu_lock);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200316 srcu_read_unlock(&kvm->srcu, idx);
Izik Eidus3da0dd42009-09-23 21:47:18 +0300317}
318
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200319static void kvm_mmu_notifier_invalidate_range_start(struct mmu_notifier *mn,
320 struct mm_struct *mm,
321 unsigned long start,
322 unsigned long end)
323{
324 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200325 int need_tlb_flush = 0, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200326
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200327 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200328 spin_lock(&kvm->mmu_lock);
329 /*
330 * The count increase must become visible at unlock time as no
331 * spte can be established without taking the mmu_lock and
332 * count is also read inside the mmu_lock critical section.
333 */
334 kvm->mmu_notifier_count++;
335 for (; start < end; start += PAGE_SIZE)
336 need_tlb_flush |= kvm_unmap_hva(kvm, start);
Xiao Guangronga4ee1ca2010-11-23 11:13:00 +0800337 need_tlb_flush |= kvm->tlbs_dirty;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200338 /* we've to flush the tlb before the pages can be freed */
339 if (need_tlb_flush)
340 kvm_flush_remote_tlbs(kvm);
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900341
342 spin_unlock(&kvm->mmu_lock);
343 srcu_read_unlock(&kvm->srcu, idx);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200344}
345
346static void kvm_mmu_notifier_invalidate_range_end(struct mmu_notifier *mn,
347 struct mm_struct *mm,
348 unsigned long start,
349 unsigned long end)
350{
351 struct kvm *kvm = mmu_notifier_to_kvm(mn);
352
353 spin_lock(&kvm->mmu_lock);
354 /*
355 * This sequence increase will notify the kvm page fault that
356 * the page that is going to be mapped in the spte could have
357 * been freed.
358 */
359 kvm->mmu_notifier_seq++;
Paul Mackerrasa355aa52011-12-12 12:37:21 +0000360 smp_wmb();
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200361 /*
362 * The above sequence increase must be visible before the
Paul Mackerrasa355aa52011-12-12 12:37:21 +0000363 * below count decrease, which is ensured by the smp_wmb above
364 * in conjunction with the smp_rmb in mmu_notifier_retry().
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200365 */
366 kvm->mmu_notifier_count--;
367 spin_unlock(&kvm->mmu_lock);
368
369 BUG_ON(kvm->mmu_notifier_count < 0);
370}
371
372static int kvm_mmu_notifier_clear_flush_young(struct mmu_notifier *mn,
373 struct mm_struct *mm,
374 unsigned long address)
375{
376 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200377 int young, idx;
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200378
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200379 idx = srcu_read_lock(&kvm->srcu);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200380 spin_lock(&kvm->mmu_lock);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200381
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900382 young = kvm_age_hva(kvm, address);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200383 if (young)
384 kvm_flush_remote_tlbs(kvm);
385
Takuya Yoshikawa565f3be2012-02-10 15:28:31 +0900386 spin_unlock(&kvm->mmu_lock);
387 srcu_read_unlock(&kvm->srcu, idx);
388
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200389 return young;
390}
391
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800392static int kvm_mmu_notifier_test_young(struct mmu_notifier *mn,
393 struct mm_struct *mm,
394 unsigned long address)
395{
396 struct kvm *kvm = mmu_notifier_to_kvm(mn);
397 int young, idx;
398
399 idx = srcu_read_lock(&kvm->srcu);
400 spin_lock(&kvm->mmu_lock);
401 young = kvm_test_age_hva(kvm, address);
402 spin_unlock(&kvm->mmu_lock);
403 srcu_read_unlock(&kvm->srcu, idx);
404
405 return young;
406}
407
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100408static void kvm_mmu_notifier_release(struct mmu_notifier *mn,
409 struct mm_struct *mm)
410{
411 struct kvm *kvm = mmu_notifier_to_kvm(mn);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800412 int idx;
413
414 idx = srcu_read_lock(&kvm->srcu);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100415 kvm_arch_flush_shadow(kvm);
Lai Jiangshaneda2bed2010-04-20 14:29:29 +0800416 srcu_read_unlock(&kvm->srcu, idx);
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100417}
418
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200419static const struct mmu_notifier_ops kvm_mmu_notifier_ops = {
420 .invalidate_page = kvm_mmu_notifier_invalidate_page,
421 .invalidate_range_start = kvm_mmu_notifier_invalidate_range_start,
422 .invalidate_range_end = kvm_mmu_notifier_invalidate_range_end,
423 .clear_flush_young = kvm_mmu_notifier_clear_flush_young,
Andrea Arcangeli8ee53822011-01-13 15:47:10 -0800424 .test_young = kvm_mmu_notifier_test_young,
Izik Eidus3da0dd42009-09-23 21:47:18 +0300425 .change_pte = kvm_mmu_notifier_change_pte,
Marcelo Tosatti85db06e2008-12-10 21:23:26 +0100426 .release = kvm_mmu_notifier_release,
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200427};
Avi Kivity4c07b0a2009-12-20 14:54:04 +0200428
429static int kvm_init_mmu_notifier(struct kvm *kvm)
430{
431 kvm->mmu_notifier.ops = &kvm_mmu_notifier_ops;
432 return mmu_notifier_register(&kvm->mmu_notifier, current->mm);
433}
434
435#else /* !(CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER) */
436
437static int kvm_init_mmu_notifier(struct kvm *kvm)
438{
439 return 0;
440}
441
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200442#endif /* CONFIG_MMU_NOTIFIER && KVM_ARCH_WANT_MMU_NOTIFIER */
443
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800444static void kvm_init_memslots_id(struct kvm *kvm)
445{
446 int i;
447 struct kvm_memslots *slots = kvm->memslots;
448
449 for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800450 slots->id_to_index[i] = slots->memslots[i].id = i;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800451}
452
Carsten Ottee08b9632012-01-04 10:25:20 +0100453static struct kvm *kvm_create_vm(unsigned long type)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800454{
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100455 int r, i;
456 struct kvm *kvm = kvm_arch_alloc_vm();
Avi Kivity6aa8b732006-12-10 02:21:36 -0800457
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100458 if (!kvm)
459 return ERR_PTR(-ENOMEM);
460
Carsten Ottee08b9632012-01-04 10:25:20 +0100461 r = kvm_arch_init_vm(kvm, type);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100462 if (r)
463 goto out_err_nodisable;
Alexander Graf10474ae2009-09-15 11:37:46 +0200464
465 r = hardware_enable_all();
466 if (r)
467 goto out_err_nodisable;
468
Avi Kivity75858a82009-01-04 17:10:50 +0200469#ifdef CONFIG_HAVE_KVM_IRQCHIP
470 INIT_HLIST_HEAD(&kvm->mask_notifier_list);
Gleb Natapov136bdfe2009-08-24 11:54:23 +0300471 INIT_HLIST_HEAD(&kvm->irq_ack_notifier_list);
Avi Kivity75858a82009-01-04 17:10:50 +0200472#endif
Avi Kivity6aa8b732006-12-10 02:21:36 -0800473
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200474 r = -ENOMEM;
475 kvm->memslots = kzalloc(sizeof(struct kvm_memslots), GFP_KERNEL);
476 if (!kvm->memslots)
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100477 goto out_err_nosrcu;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800478 kvm_init_memslots_id(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200479 if (init_srcu_struct(&kvm->srcu))
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100480 goto out_err_nosrcu;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200481 for (i = 0; i < KVM_NR_BUSES; i++) {
482 kvm->buses[i] = kzalloc(sizeof(struct kvm_io_bus),
483 GFP_KERNEL);
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100484 if (!kvm->buses[i])
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200485 goto out_err;
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200486 }
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700487
OGAWA Hirofumi85722cd2011-05-11 09:28:28 +0900488 spin_lock_init(&kvm->mmu_lock);
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200489 kvm->mm = current->mm;
490 atomic_inc(&kvm->mm->mm_count);
Gregory Haskinsd34e6b12009-07-07 17:08:49 -0400491 kvm_eventfd_init(kvm);
Shaohua Li11ec2802007-07-23 14:51:37 +0800492 mutex_init(&kvm->lock);
Marcelo Tosatti60eead72009-06-04 15:08:23 -0300493 mutex_init(&kvm->irq_lock);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200494 mutex_init(&kvm->slots_lock);
Izik Eidusd39f13b2008-03-30 16:01:25 +0300495 atomic_set(&kvm->users_count, 1);
Mike Waychison74b5c5b2011-06-03 13:04:53 -0700496
497 r = kvm_init_mmu_notifier(kvm);
498 if (r)
499 goto out_err;
500
Jan Kiszkae935b832011-02-08 12:55:33 +0100501 raw_spin_lock(&kvm_lock);
Rusty Russell5e58cfe2007-07-23 17:08:21 +1000502 list_add(&kvm->vm_list, &vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100503 raw_spin_unlock(&kvm_lock);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100504
Avi Kivityf17abe92007-02-21 19:28:04 +0200505 return kvm;
Alexander Graf10474ae2009-09-15 11:37:46 +0200506
507out_err:
Jan Kiszka57e7fbe2010-11-09 12:42:12 +0100508 cleanup_srcu_struct(&kvm->srcu);
509out_err_nosrcu:
Alexander Graf10474ae2009-09-15 11:37:46 +0200510 hardware_disable_all();
511out_err_nodisable:
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200512 for (i = 0; i < KVM_NR_BUSES; i++)
513 kfree(kvm->buses[i]);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200514 kfree(kvm->memslots);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100515 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200516 return ERR_PTR(r);
Avi Kivityf17abe92007-02-21 19:28:04 +0200517}
518
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900519static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
520{
521 if (!memslot->dirty_bitmap)
522 return;
523
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900524 if (2 * kvm_dirty_bitmap_bytes(memslot) > PAGE_SIZE)
525 vfree(memslot->dirty_bitmap_head);
526 else
527 kfree(memslot->dirty_bitmap_head);
528
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900529 memslot->dirty_bitmap = NULL;
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900530 memslot->dirty_bitmap_head = NULL;
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900531}
532
Avi Kivity6aa8b732006-12-10 02:21:36 -0800533/*
534 * Free any memory in @free but not in @dont.
535 */
536static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
537 struct kvm_memory_slot *dont)
538{
Izik Eidus290fc382007-09-27 14:11:22 +0200539 if (!dont || free->rmap != dont->rmap)
540 vfree(free->rmap);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800541
542 if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900543 kvm_destroy_dirty_bitmap(free);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800544
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900545 kvm_arch_free_memslot(free, dont);
Marcelo Tosatti05da4552008-02-23 11:44:30 -0300546
Avi Kivity6aa8b732006-12-10 02:21:36 -0800547 free->npages = 0;
Anthony Liguori8d4e1282007-10-18 09:59:34 -0500548 free->rmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800549}
550
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800551void kvm_free_physmem(struct kvm *kvm)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800552{
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200553 struct kvm_memslots *slots = kvm->memslots;
Xiao Guangrongbe6ba0f2011-11-24 17:39:18 +0800554 struct kvm_memory_slot *memslot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800555
Xiao Guangrongbe6ba0f2011-11-24 17:39:18 +0800556 kvm_for_each_memslot(memslot, slots)
557 kvm_free_physmem_slot(memslot, NULL);
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200558
559 kfree(kvm->memslots);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800560}
561
Avi Kivityf17abe92007-02-21 19:28:04 +0200562static void kvm_destroy_vm(struct kvm *kvm)
563{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200564 int i;
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200565 struct mm_struct *mm = kvm->mm;
566
Sheng Yangad8ba2c2009-01-06 10:03:02 +0800567 kvm_arch_sync_events(kvm);
Jan Kiszkae935b832011-02-08 12:55:33 +0100568 raw_spin_lock(&kvm_lock);
Avi Kivity133de902007-02-12 00:54:44 -0800569 list_del(&kvm->vm_list);
Jan Kiszkae935b832011-02-08 12:55:33 +0100570 raw_spin_unlock(&kvm_lock);
Avi Kivity399ec802008-11-19 13:58:46 +0200571 kvm_free_irq_routing(kvm);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -0200572 for (i = 0; i < KVM_NR_BUSES; i++)
573 kvm_io_bus_destroy(kvm->buses[i]);
Avi Kivity980da6c2009-12-20 15:13:43 +0200574 kvm_coalesced_mmio_free(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200575#if defined(CONFIG_MMU_NOTIFIER) && defined(KVM_ARCH_WANT_MMU_NOTIFIER)
576 mmu_notifier_unregister(&kvm->mmu_notifier, kvm->mm);
Gleb Natapovf00be0c2009-03-19 12:20:36 +0200577#else
578 kvm_arch_flush_shadow(kvm);
Andrea Arcangelie930bff2008-07-25 16:24:52 +0200579#endif
Zhang Xiantaod19a9cd2007-11-18 18:43:45 +0800580 kvm_arch_destroy_vm(kvm);
Jan Kiszkad89f5ef2010-11-09 17:02:49 +0100581 kvm_free_physmem(kvm);
582 cleanup_srcu_struct(&kvm->srcu);
583 kvm_arch_free_vm(kvm);
Alexander Graf10474ae2009-09-15 11:37:46 +0200584 hardware_disable_all();
Avi Kivity6d4e4c42007-11-21 16:41:05 +0200585 mmdrop(mm);
Avi Kivityf17abe92007-02-21 19:28:04 +0200586}
587
Izik Eidusd39f13b2008-03-30 16:01:25 +0300588void kvm_get_kvm(struct kvm *kvm)
589{
590 atomic_inc(&kvm->users_count);
591}
592EXPORT_SYMBOL_GPL(kvm_get_kvm);
593
594void kvm_put_kvm(struct kvm *kvm)
595{
596 if (atomic_dec_and_test(&kvm->users_count))
597 kvm_destroy_vm(kvm);
598}
599EXPORT_SYMBOL_GPL(kvm_put_kvm);
600
601
Avi Kivityf17abe92007-02-21 19:28:04 +0200602static int kvm_vm_release(struct inode *inode, struct file *filp)
603{
604 struct kvm *kvm = filp->private_data;
605
Gregory Haskins721eecb2009-05-20 10:30:49 -0400606 kvm_irqfd_release(kvm);
607
Izik Eidusd39f13b2008-03-30 16:01:25 +0300608 kvm_put_kvm(kvm);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800609 return 0;
610}
611
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900612/*
613 * Allocation size is twice as large as the actual dirty bitmap size.
614 * This makes it possible to do double buffering: see x86's
615 * kvm_vm_ioctl_get_dirty_log().
616 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900617static int kvm_create_dirty_bitmap(struct kvm_memory_slot *memslot)
618{
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900619#ifndef CONFIG_S390
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900620 unsigned long dirty_bytes = 2 * kvm_dirty_bitmap_bytes(memslot);
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900621
Takuya Yoshikawa6f9e5c172010-11-01 14:36:09 +0900622 if (dirty_bytes > PAGE_SIZE)
623 memslot->dirty_bitmap = vzalloc(dirty_bytes);
624 else
625 memslot->dirty_bitmap = kzalloc(dirty_bytes, GFP_KERNEL);
626
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900627 if (!memslot->dirty_bitmap)
628 return -ENOMEM;
629
Takuya Yoshikawa515a0122010-10-27 18:23:54 +0900630 memslot->dirty_bitmap_head = memslot->dirty_bitmap;
Takuya Yoshikawa7850ac52011-11-14 18:23:34 +0900631 memslot->nr_dirty_pages = 0;
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900632#endif /* !CONFIG_S390 */
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900633 return 0;
634}
635
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800636static int cmp_memslot(const void *slot1, const void *slot2)
637{
638 struct kvm_memory_slot *s1, *s2;
639
640 s1 = (struct kvm_memory_slot *)slot1;
641 s2 = (struct kvm_memory_slot *)slot2;
642
643 if (s1->npages < s2->npages)
644 return 1;
645 if (s1->npages > s2->npages)
646 return -1;
647
648 return 0;
649}
650
651/*
652 * Sort the memslots base on its size, so the larger slots
653 * will get better fit.
654 */
655static void sort_memslots(struct kvm_memslots *slots)
656{
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800657 int i;
658
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800659 sort(slots->memslots, KVM_MEM_SLOTS_NUM,
660 sizeof(struct kvm_memory_slot), cmp_memslot, NULL);
Xiao Guangrongf85e2cb2011-11-24 17:41:54 +0800661
662 for (i = 0; i < KVM_MEM_SLOTS_NUM; i++)
663 slots->id_to_index[slots->memslots[i].id] = i;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800664}
665
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800666void update_memslots(struct kvm_memslots *slots, struct kvm_memory_slot *new)
667{
668 if (new) {
669 int id = new->id;
Xiao Guangrong28a37542011-11-24 19:04:35 +0800670 struct kvm_memory_slot *old = id_to_memslot(slots, id);
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800671 unsigned long npages = old->npages;
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800672
Xiao Guangrong28a37542011-11-24 19:04:35 +0800673 *old = *new;
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800674 if (new->npages != npages)
675 sort_memslots(slots);
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800676 }
677
678 slots->generation++;
679}
680
Avi Kivity6aa8b732006-12-10 02:21:36 -0800681/*
Avi Kivity6aa8b732006-12-10 02:21:36 -0800682 * Allocate some memory and give it an address in the guest physical address
683 * space.
684 *
685 * Discontiguous memory is allowed, mostly for framebuffers.
Sheng Yangf78e0e22007-10-29 09:40:42 +0800686 *
Marcelo Tosatti10589a42007-12-20 19:18:22 -0500687 * Must be called holding mmap_sem for write.
Avi Kivity6aa8b732006-12-10 02:21:36 -0800688 */
Sheng Yangf78e0e22007-10-29 09:40:42 +0800689int __kvm_set_memory_region(struct kvm *kvm,
690 struct kvm_userspace_memory_region *mem,
691 int user_alloc)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800692{
Avi Kivity8234b222010-12-27 12:08:45 +0200693 int r;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800694 gfn_t base_gfn;
Heiko Carstens28bcb112009-09-03 17:35:35 +0200695 unsigned long npages;
696 unsigned long i;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800697 struct kvm_memory_slot *memslot;
698 struct kvm_memory_slot old, new;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200699 struct kvm_memslots *slots, *old_memslots;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800700
701 r = -EINVAL;
702 /* General sanity checks */
703 if (mem->memory_size & (PAGE_SIZE - 1))
704 goto out;
705 if (mem->guest_phys_addr & (PAGE_SIZE - 1))
706 goto out;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +0900707 /* We can read the guest memory with __xxx_user() later on. */
708 if (user_alloc &&
709 ((mem->userspace_addr & (PAGE_SIZE - 1)) ||
Heiko Carstens9e3bb6b2011-05-24 07:51:27 +0200710 !access_ok(VERIFY_WRITE,
711 (void __user *)(unsigned long)mem->userspace_addr,
712 mem->memory_size)))
Hollis Blanchard78749802008-11-07 13:32:12 -0600713 goto out;
Xiao Guangrong93a5cef2011-11-24 17:37:48 +0800714 if (mem->slot >= KVM_MEM_SLOTS_NUM)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800715 goto out;
716 if (mem->guest_phys_addr + mem->memory_size < mem->guest_phys_addr)
717 goto out;
718
Xiao Guangrong28a37542011-11-24 19:04:35 +0800719 memslot = id_to_memslot(kvm->memslots, mem->slot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800720 base_gfn = mem->guest_phys_addr >> PAGE_SHIFT;
721 npages = mem->memory_size >> PAGE_SHIFT;
722
Takuya Yoshikawa660c22c2010-04-13 22:47:24 +0900723 r = -EINVAL;
724 if (npages > KVM_MEM_MAX_NR_PAGES)
725 goto out;
726
Avi Kivity6aa8b732006-12-10 02:21:36 -0800727 if (!npages)
728 mem->flags &= ~KVM_MEM_LOG_DIRTY_PAGES;
729
Avi Kivity6aa8b732006-12-10 02:21:36 -0800730 new = old = *memslot;
731
Avi Kivitye36d96f2010-06-21 10:56:36 +0300732 new.id = mem->slot;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800733 new.base_gfn = base_gfn;
734 new.npages = npages;
735 new.flags = mem->flags;
736
737 /* Disallow changing a memory slot's size. */
738 r = -EINVAL;
739 if (npages && old.npages && npages != old.npages)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800740 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800741
742 /* Check for overlaps */
743 r = -EEXIST;
744 for (i = 0; i < KVM_MEMORY_SLOTS; ++i) {
Marcelo Tosatti46a26bf2009-12-23 14:35:16 -0200745 struct kvm_memory_slot *s = &kvm->memslots->memslots[i];
Avi Kivity6aa8b732006-12-10 02:21:36 -0800746
Jan Kiszka4cd481f2009-04-13 11:59:32 +0200747 if (s == memslot || !s->npages)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800748 continue;
749 if (!((base_gfn + npages <= s->base_gfn) ||
750 (base_gfn >= s->base_gfn + s->npages)))
Sheng Yangf78e0e22007-10-29 09:40:42 +0800751 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800752 }
Avi Kivity6aa8b732006-12-10 02:21:36 -0800753
Avi Kivity6aa8b732006-12-10 02:21:36 -0800754 /* Free page dirty bitmap if unneeded */
755 if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
Al Viro8b6d44c2007-02-09 16:38:40 +0000756 new.dirty_bitmap = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800757
758 r = -ENOMEM;
759
760 /* Allocate if a slot is being created */
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900761 if (npages && !old.npages) {
762 new.user_alloc = user_alloc;
763 new.userspace_addr = mem->userspace_addr;
Carsten Otteeff01142008-06-27 15:05:31 +0200764#ifndef CONFIG_S390
Takuya Yoshikawa26535032010-11-02 10:49:34 +0900765 new.rmap = vzalloc(npages * sizeof(*new.rmap));
Izik Eidus290fc382007-09-27 14:11:22 +0200766 if (!new.rmap)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800767 goto out_free;
Takuya Yoshikawa189a2f72012-02-08 13:01:09 +0900768#endif /* not defined CONFIG_S390 */
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900769 if (kvm_arch_create_memslot(&new, npages))
770 goto out_free;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800771 }
Joerg Roedelec04b262009-06-19 15:16:23 +0200772
Avi Kivity6aa8b732006-12-10 02:21:36 -0800773 /* Allocate page dirty bitmap if needed */
774 if ((new.flags & KVM_MEM_LOG_DIRTY_PAGES) && !new.dirty_bitmap) {
Takuya Yoshikawaa36a57b12010-10-27 18:22:19 +0900775 if (kvm_create_dirty_bitmap(&new) < 0)
Sheng Yangf78e0e22007-10-29 09:40:42 +0800776 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200777 /* destroy any largepage mappings for dirty tracking */
Avi Kivity6aa8b732006-12-10 02:21:36 -0800778 }
779
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200780 if (!npages) {
Xiao Guangrong28a37542011-11-24 19:04:35 +0800781 struct kvm_memory_slot *slot;
782
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200783 r = -ENOMEM;
Thomas Meyer6da64fd2011-11-08 20:32:19 +0100784 slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots),
785 GFP_KERNEL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200786 if (!slots)
787 goto out_free;
Xiao Guangrong28a37542011-11-24 19:04:35 +0800788 slot = id_to_memslot(slots, mem->slot);
789 slot->flags |= KVM_MEMSLOT_INVALID;
790
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800791 update_memslots(slots, NULL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200792
793 old_memslots = kvm->memslots;
794 rcu_assign_pointer(kvm->memslots, slots);
795 synchronize_srcu_expedited(&kvm->srcu);
796 /* From this point no new shadow pages pointing to a deleted
797 * memslot will be created.
798 *
799 * validation of sp->gfn happens in:
800 * - gfn_to_hva (kvm_read_guest, gfn_to_pfn)
801 * - kvm_is_visible_gfn (mmu_check_roots)
802 */
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300803 kvm_arch_flush_shadow(kvm);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200804 kfree(old_memslots);
805 }
Marcelo Tosatti34d4cb82008-07-10 20:49:31 -0300806
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200807 r = kvm_arch_prepare_memory_region(kvm, &new, old, mem, user_alloc);
808 if (r)
809 goto out_free;
810
Alex Williamson32f6daa2012-04-11 09:51:49 -0600811 /* map/unmap the pages in iommu page table */
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200812 if (npages) {
813 r = kvm_iommu_map_pages(kvm, &new);
814 if (r)
815 goto out_free;
Alex Williamson32f6daa2012-04-11 09:51:49 -0600816 } else
817 kvm_iommu_unmap_pages(kvm, &old);
Andrea Arcangeli604b38a2008-07-25 16:32:03 +0200818
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200819 r = -ENOMEM;
Thomas Meyer6da64fd2011-11-08 20:32:19 +0100820 slots = kmemdup(kvm->memslots, sizeof(struct kvm_memslots),
821 GFP_KERNEL);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200822 if (!slots)
823 goto out_free;
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200824
825 /* actual memory is freed via old in kvm_free_physmem_slot below */
826 if (!npages) {
827 new.rmap = NULL;
828 new.dirty_bitmap = NULL;
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900829 memset(&new.arch, 0, sizeof(new.arch));
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200830 }
831
Xiao Guangrongbe593d62011-11-24 17:38:24 +0800832 update_memslots(slots, &new);
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200833 old_memslots = kvm->memslots;
834 rcu_assign_pointer(kvm->memslots, slots);
835 synchronize_srcu_expedited(&kvm->srcu);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800836
Marcelo Tosattif7784b82009-12-23 14:35:18 -0200837 kvm_arch_commit_memory_region(kvm, mem, old, user_alloc);
Zhang Xiantao3ad82a72007-11-20 13:11:38 +0800838
Xiao Guangrongce88dec2011-07-12 03:33:44 +0800839 /*
840 * If the new memory slot is created, we need to clear all
841 * mmio sptes.
842 */
843 if (npages && old.base_gfn != mem->guest_phys_addr >> PAGE_SHIFT)
844 kvm_arch_flush_shadow(kvm);
845
Marcelo Tosattibc6678a2009-12-23 14:35:21 -0200846 kvm_free_physmem_slot(&old, &new);
847 kfree(old_memslots);
848
Avi Kivity6aa8b732006-12-10 02:21:36 -0800849 return 0;
850
Sheng Yangf78e0e22007-10-29 09:40:42 +0800851out_free:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800852 kvm_free_physmem_slot(&new, &old);
853out:
854 return r;
Izik Eidus210c7c42007-10-24 23:52:57 +0200855
856}
Sheng Yangf78e0e22007-10-29 09:40:42 +0800857EXPORT_SYMBOL_GPL(__kvm_set_memory_region);
858
859int kvm_set_memory_region(struct kvm *kvm,
860 struct kvm_userspace_memory_region *mem,
861 int user_alloc)
862{
863 int r;
864
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200865 mutex_lock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800866 r = __kvm_set_memory_region(kvm, mem, user_alloc);
Marcelo Tosatti79fac952009-12-23 14:35:26 -0200867 mutex_unlock(&kvm->slots_lock);
Sheng Yangf78e0e22007-10-29 09:40:42 +0800868 return r;
869}
Izik Eidus210c7c42007-10-24 23:52:57 +0200870EXPORT_SYMBOL_GPL(kvm_set_memory_region);
871
Carsten Otte1fe779f2007-10-29 16:08:35 +0100872int kvm_vm_ioctl_set_memory_region(struct kvm *kvm,
873 struct
874 kvm_userspace_memory_region *mem,
875 int user_alloc)
Izik Eidus210c7c42007-10-24 23:52:57 +0200876{
Izik Eiduse0d62c72007-10-24 23:57:46 +0200877 if (mem->slot >= KVM_MEMORY_SLOTS)
878 return -EINVAL;
Izik Eidus210c7c42007-10-24 23:52:57 +0200879 return kvm_set_memory_region(kvm, mem, user_alloc);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800880}
881
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800882int kvm_get_dirty_log(struct kvm *kvm,
883 struct kvm_dirty_log *log, int *is_dirty)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800884{
885 struct kvm_memory_slot *memslot;
886 int r, i;
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900887 unsigned long n;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800888 unsigned long any = 0;
889
Avi Kivity6aa8b732006-12-10 02:21:36 -0800890 r = -EINVAL;
891 if (log->slot >= KVM_MEMORY_SLOTS)
892 goto out;
893
Xiao Guangrong28a37542011-11-24 19:04:35 +0800894 memslot = id_to_memslot(kvm->memslots, log->slot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800895 r = -ENOENT;
896 if (!memslot->dirty_bitmap)
897 goto out;
898
Takuya Yoshikawa87bf6e72010-04-12 19:35:35 +0900899 n = kvm_dirty_bitmap_bytes(memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800900
Uri Lublincd1a4a92007-02-22 16:43:09 +0200901 for (i = 0; !any && i < n/sizeof(long); ++i)
Avi Kivity6aa8b732006-12-10 02:21:36 -0800902 any = memslot->dirty_bitmap[i];
903
904 r = -EFAULT;
905 if (copy_to_user(log->dirty_bitmap, memslot->dirty_bitmap, n))
906 goto out;
907
Zhang Xiantao5bb064d2007-11-18 20:29:43 +0800908 if (any)
909 *is_dirty = 1;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800910
911 r = 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -0800912out:
Avi Kivity6aa8b732006-12-10 02:21:36 -0800913 return r;
914}
915
Takuya Yoshikawadb3fe4e2012-02-08 13:02:18 +0900916bool kvm_largepages_enabled(void)
917{
918 return largepages_enabled;
919}
920
Marcelo Tosatti54dee992009-06-11 12:07:44 -0300921void kvm_disable_largepages(void)
922{
923 largepages_enabled = false;
924}
925EXPORT_SYMBOL_GPL(kvm_disable_largepages);
926
Izik Eiduscea7bb22007-10-17 19:17:48 +0200927int is_error_page(struct page *page)
928{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300929 return page == bad_page || page == hwpoison_page || page == fault_page;
Izik Eiduscea7bb22007-10-17 19:17:48 +0200930}
931EXPORT_SYMBOL_GPL(is_error_page);
932
Anthony Liguori35149e22008-04-02 14:46:56 -0500933int is_error_pfn(pfn_t pfn)
934{
Gleb Natapovedba23e2010-07-07 20:16:45 +0300935 return pfn == bad_pfn || pfn == hwpoison_pfn || pfn == fault_pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -0500936}
937EXPORT_SYMBOL_GPL(is_error_pfn);
938
Huang Yingbf998152010-05-31 14:28:19 +0800939int is_hwpoison_pfn(pfn_t pfn)
940{
941 return pfn == hwpoison_pfn;
942}
943EXPORT_SYMBOL_GPL(is_hwpoison_pfn);
944
Gleb Natapovedba23e2010-07-07 20:16:45 +0300945int is_fault_pfn(pfn_t pfn)
946{
947 return pfn == fault_pfn;
948}
949EXPORT_SYMBOL_GPL(is_fault_pfn);
950
Xiao Guangrongfce92dc2011-07-12 03:28:54 +0800951int is_noslot_pfn(pfn_t pfn)
952{
953 return pfn == bad_pfn;
954}
955EXPORT_SYMBOL_GPL(is_noslot_pfn);
956
957int is_invalid_pfn(pfn_t pfn)
958{
959 return pfn == hwpoison_pfn || pfn == fault_pfn;
960}
961EXPORT_SYMBOL_GPL(is_invalid_pfn);
962
Izik Eidusf9d46eb2007-11-11 22:02:22 +0200963static inline unsigned long bad_hva(void)
964{
965 return PAGE_OFFSET;
966}
967
968int kvm_is_error_hva(unsigned long addr)
969{
970 return addr == bad_hva();
971}
972EXPORT_SYMBOL_GPL(kvm_is_error_hva);
973
Gleb Natapov49c77542010-10-18 15:22:23 +0200974struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
975{
976 return __gfn_to_memslot(kvm_memslots(kvm), gfn);
977}
Avi Kivitya1f4d3952010-06-21 11:44:20 +0300978EXPORT_SYMBOL_GPL(gfn_to_memslot);
Avi Kivity6aa8b732006-12-10 02:21:36 -0800979
Izik Eiduse0d62c72007-10-24 23:57:46 +0200980int kvm_is_visible_gfn(struct kvm *kvm, gfn_t gfn)
981{
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800982 struct kvm_memory_slot *memslot = gfn_to_memslot(kvm, gfn);
Izik Eiduse0d62c72007-10-24 23:57:46 +0200983
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800984 if (!memslot || memslot->id >= KVM_MEMORY_SLOTS ||
985 memslot->flags & KVM_MEMSLOT_INVALID)
986 return 0;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200987
Xiao Guangrongbf3e05b2011-11-24 17:40:57 +0800988 return 1;
Izik Eiduse0d62c72007-10-24 23:57:46 +0200989}
990EXPORT_SYMBOL_GPL(kvm_is_visible_gfn);
991
Joerg Roedel8f0b1ab2010-01-28 12:37:56 +0100992unsigned long kvm_host_page_size(struct kvm *kvm, gfn_t gfn)
993{
994 struct vm_area_struct *vma;
995 unsigned long addr, size;
996
997 size = PAGE_SIZE;
998
999 addr = gfn_to_hva(kvm, gfn);
1000 if (kvm_is_error_hva(addr))
1001 return PAGE_SIZE;
1002
1003 down_read(&current->mm->mmap_sem);
1004 vma = find_vma(current->mm, addr);
1005 if (!vma)
1006 goto out;
1007
1008 size = vma_kernel_pagesize(vma);
1009
1010out:
1011 up_read(&current->mm->mmap_sem);
1012
1013 return size;
1014}
1015
Gleb Natapov49c77542010-10-18 15:22:23 +02001016static unsigned long gfn_to_hva_many(struct kvm_memory_slot *slot, gfn_t gfn,
Xiao Guangrong48987782010-08-22 19:11:43 +08001017 gfn_t *nr_pages)
Izik Eidus539cb662007-11-11 22:05:04 +02001018{
Marcelo Tosattibc6678a2009-12-23 14:35:21 -02001019 if (!slot || slot->flags & KVM_MEMSLOT_INVALID)
Izik Eidus539cb662007-11-11 22:05:04 +02001020 return bad_hva();
Xiao Guangrong48987782010-08-22 19:11:43 +08001021
1022 if (nr_pages)
1023 *nr_pages = slot->npages - (gfn - slot->base_gfn);
1024
Takuya Yoshikawaf5c98032010-02-25 11:33:19 +09001025 return gfn_to_hva_memslot(slot, gfn);
Izik Eidus539cb662007-11-11 22:05:04 +02001026}
Xiao Guangrong48987782010-08-22 19:11:43 +08001027
1028unsigned long gfn_to_hva(struct kvm *kvm, gfn_t gfn)
1029{
Gleb Natapov49c77542010-10-18 15:22:23 +02001030 return gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, NULL);
Xiao Guangrong48987782010-08-22 19:11:43 +08001031}
Sheng Yang0d150292008-04-25 21:44:50 +08001032EXPORT_SYMBOL_GPL(gfn_to_hva);
Izik Eidus539cb662007-11-11 22:05:04 +02001033
Gleb Natapov80300892010-10-19 18:13:41 +02001034static pfn_t get_fault_pfn(void)
1035{
1036 get_page(fault_page);
1037 return fault_pfn;
1038}
1039
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001040int get_user_page_nowait(struct task_struct *tsk, struct mm_struct *mm,
1041 unsigned long start, int write, struct page **page)
1042{
1043 int flags = FOLL_TOUCH | FOLL_NOWAIT | FOLL_HWPOISON | FOLL_GET;
1044
1045 if (write)
1046 flags |= FOLL_WRITE;
1047
1048 return __get_user_pages(tsk, mm, start, 1, flags, page, NULL, NULL);
1049}
1050
Huang Yingfafc3db2011-01-30 11:15:49 +08001051static inline int check_user_page_hwpoison(unsigned long addr)
1052{
1053 int rc, flags = FOLL_TOUCH | FOLL_HWPOISON | FOLL_WRITE;
1054
1055 rc = __get_user_pages(current, current->mm, addr, 1,
1056 flags, NULL, NULL, NULL);
1057 return rc == -EHWPOISON;
1058}
1059
Gleb Natapovaf585b92010-10-14 11:22:46 +02001060static pfn_t hva_to_pfn(struct kvm *kvm, unsigned long addr, bool atomic,
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001061 bool *async, bool write_fault, bool *writable)
Avi Kivity954bbbc2007-03-30 14:02:32 +03001062{
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001063 struct page *page[1];
Gleb Natapovaf585b92010-10-14 11:22:46 +02001064 int npages = 0;
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001065 pfn_t pfn;
Avi Kivity954bbbc2007-03-30 14:02:32 +03001066
Gleb Natapovaf585b92010-10-14 11:22:46 +02001067 /* we can do it either atomically or asynchronously, not both */
1068 BUG_ON(atomic && async);
1069
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001070 BUG_ON(!write_fault && !writable);
1071
1072 if (writable)
1073 *writable = true;
1074
Gleb Natapovaf585b92010-10-14 11:22:46 +02001075 if (atomic || async)
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001076 npages = __get_user_pages_fast(addr, 1, 1, page);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001077
1078 if (unlikely(npages != 1) && !atomic) {
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001079 might_sleep();
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001080
1081 if (writable)
1082 *writable = write_fault;
1083
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001084 if (async) {
1085 down_read(&current->mm->mmap_sem);
1086 npages = get_user_page_nowait(current, current->mm,
1087 addr, write_fault, page);
1088 up_read(&current->mm->mmap_sem);
1089 } else
1090 npages = get_user_pages_fast(addr, 1, write_fault,
1091 page);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001092
1093 /* map read fault as writable if possible */
1094 if (unlikely(!write_fault) && npages == 1) {
1095 struct page *wpage[1];
1096
1097 npages = __get_user_pages_fast(addr, 1, 1, wpage);
1098 if (npages == 1) {
1099 *writable = true;
1100 put_page(page[0]);
1101 page[0] = wpage[0];
1102 }
1103 npages = 1;
1104 }
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001105 }
Izik Eidus539cb662007-11-11 22:05:04 +02001106
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001107 if (unlikely(npages != 1)) {
1108 struct vm_area_struct *vma;
Anthony Liguori8d4e1282007-10-18 09:59:34 -05001109
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001110 if (atomic)
Gleb Natapov80300892010-10-19 18:13:41 +02001111 return get_fault_pfn();
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001112
Huang Yingbbeb3402010-06-22 14:23:11 +08001113 down_read(&current->mm->mmap_sem);
Gleb Natapov0857b9e2011-02-01 13:21:47 +02001114 if (npages == -EHWPOISON ||
1115 (!async && check_user_page_hwpoison(addr))) {
Huang Yingbbeb3402010-06-22 14:23:11 +08001116 up_read(&current->mm->mmap_sem);
Huang Yingbf998152010-05-31 14:28:19 +08001117 get_page(hwpoison_page);
1118 return page_to_pfn(hwpoison_page);
1119 }
1120
Gleb Natapov80300892010-10-19 18:13:41 +02001121 vma = find_vma_intersection(current->mm, addr, addr+1);
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001122
Gleb Natapov80300892010-10-19 18:13:41 +02001123 if (vma == NULL)
1124 pfn = get_fault_pfn();
1125 else if ((vma->vm_flags & VM_PFNMAP)) {
1126 pfn = ((addr - vma->vm_start) >> PAGE_SHIFT) +
1127 vma->vm_pgoff;
1128 BUG_ON(!kvm_is_mmio_pfn(pfn));
1129 } else {
1130 if (async && (vma->vm_flags & VM_WRITE))
Gleb Natapovaf585b92010-10-14 11:22:46 +02001131 *async = true;
Gleb Natapov80300892010-10-19 18:13:41 +02001132 pfn = get_fault_pfn();
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001133 }
Marcelo Tosatti4c2155c2008-09-16 20:54:47 -03001134 up_read(&current->mm->mmap_sem);
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001135 } else
1136 pfn = page_to_pfn(page[0]);
1137
1138 return pfn;
Anthony Liguori35149e22008-04-02 14:46:56 -05001139}
1140
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001141pfn_t hva_to_pfn_atomic(struct kvm *kvm, unsigned long addr)
1142{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001143 return hva_to_pfn(kvm, addr, true, NULL, true, NULL);
Xiao Guangrong887c08a2010-08-22 19:10:28 +08001144}
1145EXPORT_SYMBOL_GPL(hva_to_pfn_atomic);
1146
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001147static pfn_t __gfn_to_pfn(struct kvm *kvm, gfn_t gfn, bool atomic, bool *async,
1148 bool write_fault, bool *writable)
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001149{
1150 unsigned long addr;
1151
Gleb Natapovaf585b92010-10-14 11:22:46 +02001152 if (async)
1153 *async = false;
1154
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001155 addr = gfn_to_hva(kvm, gfn);
1156 if (kvm_is_error_hva(addr)) {
1157 get_page(bad_page);
1158 return page_to_pfn(bad_page);
1159 }
1160
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001161 return hva_to_pfn(kvm, addr, atomic, async, write_fault, writable);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001162}
1163
1164pfn_t gfn_to_pfn_atomic(struct kvm *kvm, gfn_t gfn)
1165{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001166 return __gfn_to_pfn(kvm, gfn, true, NULL, true, NULL);
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001167}
1168EXPORT_SYMBOL_GPL(gfn_to_pfn_atomic);
1169
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001170pfn_t gfn_to_pfn_async(struct kvm *kvm, gfn_t gfn, bool *async,
1171 bool write_fault, bool *writable)
Gleb Natapovaf585b92010-10-14 11:22:46 +02001172{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001173 return __gfn_to_pfn(kvm, gfn, false, async, write_fault, writable);
Gleb Natapovaf585b92010-10-14 11:22:46 +02001174}
1175EXPORT_SYMBOL_GPL(gfn_to_pfn_async);
1176
Xiao Guangrong365fb3f2010-08-28 19:24:13 +08001177pfn_t gfn_to_pfn(struct kvm *kvm, gfn_t gfn)
1178{
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001179 return __gfn_to_pfn(kvm, gfn, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001180}
Anthony Liguori35149e22008-04-02 14:46:56 -05001181EXPORT_SYMBOL_GPL(gfn_to_pfn);
1182
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001183pfn_t gfn_to_pfn_prot(struct kvm *kvm, gfn_t gfn, bool write_fault,
1184 bool *writable)
1185{
1186 return __gfn_to_pfn(kvm, gfn, false, NULL, write_fault, writable);
1187}
1188EXPORT_SYMBOL_GPL(gfn_to_pfn_prot);
1189
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001190pfn_t gfn_to_pfn_memslot(struct kvm *kvm,
1191 struct kvm_memory_slot *slot, gfn_t gfn)
1192{
1193 unsigned long addr = gfn_to_hva_memslot(slot, gfn);
Marcelo Tosatti612819c2010-10-22 14:18:18 -02001194 return hva_to_pfn(kvm, addr, false, NULL, true, NULL);
Marcelo Tosatti506f0d62009-12-23 14:35:19 -02001195}
1196
Xiao Guangrong48987782010-08-22 19:11:43 +08001197int gfn_to_page_many_atomic(struct kvm *kvm, gfn_t gfn, struct page **pages,
1198 int nr_pages)
1199{
1200 unsigned long addr;
1201 gfn_t entry;
1202
Gleb Natapov49c77542010-10-18 15:22:23 +02001203 addr = gfn_to_hva_many(gfn_to_memslot(kvm, gfn), gfn, &entry);
Xiao Guangrong48987782010-08-22 19:11:43 +08001204 if (kvm_is_error_hva(addr))
1205 return -1;
1206
1207 if (entry < nr_pages)
1208 return 0;
1209
1210 return __get_user_pages_fast(addr, nr_pages, 1, pages);
1211}
1212EXPORT_SYMBOL_GPL(gfn_to_page_many_atomic);
1213
Anthony Liguori35149e22008-04-02 14:46:56 -05001214struct page *gfn_to_page(struct kvm *kvm, gfn_t gfn)
1215{
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001216 pfn_t pfn;
1217
1218 pfn = gfn_to_pfn(kvm, gfn);
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001219 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001220 return pfn_to_page(pfn);
1221
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001222 WARN_ON(kvm_is_mmio_pfn(pfn));
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001223
1224 get_page(bad_page);
1225 return bad_page;
Avi Kivity954bbbc2007-03-30 14:02:32 +03001226}
Anthony Liguoriaab61cc2007-10-29 15:15:20 -05001227
Avi Kivity954bbbc2007-03-30 14:02:32 +03001228EXPORT_SYMBOL_GPL(gfn_to_page);
1229
Izik Eidusb4231d62007-11-20 11:49:33 +02001230void kvm_release_page_clean(struct page *page)
1231{
Anthony Liguori35149e22008-04-02 14:46:56 -05001232 kvm_release_pfn_clean(page_to_pfn(page));
Izik Eidusb4231d62007-11-20 11:49:33 +02001233}
1234EXPORT_SYMBOL_GPL(kvm_release_page_clean);
1235
Anthony Liguori35149e22008-04-02 14:46:56 -05001236void kvm_release_pfn_clean(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 put_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001240}
1241EXPORT_SYMBOL_GPL(kvm_release_pfn_clean);
1242
Izik Eidusb4231d62007-11-20 11:49:33 +02001243void kvm_release_page_dirty(struct page *page)
Izik Eidus8a7ae052007-10-18 11:09:33 +02001244{
Anthony Liguori35149e22008-04-02 14:46:56 -05001245 kvm_release_pfn_dirty(page_to_pfn(page));
Izik Eidus8a7ae052007-10-18 11:09:33 +02001246}
Izik Eidusb4231d62007-11-20 11:49:33 +02001247EXPORT_SYMBOL_GPL(kvm_release_page_dirty);
Izik Eidus8a7ae052007-10-18 11:09:33 +02001248
Anthony Liguori35149e22008-04-02 14:46:56 -05001249void kvm_release_pfn_dirty(pfn_t pfn)
1250{
1251 kvm_set_pfn_dirty(pfn);
1252 kvm_release_pfn_clean(pfn);
1253}
1254EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
1255
1256void kvm_set_page_dirty(struct page *page)
1257{
1258 kvm_set_pfn_dirty(page_to_pfn(page));
1259}
1260EXPORT_SYMBOL_GPL(kvm_set_page_dirty);
1261
1262void kvm_set_pfn_dirty(pfn_t pfn)
1263{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001264 if (!kvm_is_mmio_pfn(pfn)) {
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001265 struct page *page = pfn_to_page(pfn);
1266 if (!PageReserved(page))
1267 SetPageDirty(page);
1268 }
Anthony Liguori35149e22008-04-02 14:46:56 -05001269}
1270EXPORT_SYMBOL_GPL(kvm_set_pfn_dirty);
1271
1272void kvm_set_pfn_accessed(pfn_t pfn)
1273{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001274 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001275 mark_page_accessed(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001276}
1277EXPORT_SYMBOL_GPL(kvm_set_pfn_accessed);
1278
1279void kvm_get_pfn(pfn_t pfn)
1280{
Xiantao Zhangc77fb9d2008-09-27 10:55:40 +08001281 if (!kvm_is_mmio_pfn(pfn))
Anthony Liguori2e2e3732008-04-30 15:37:07 -05001282 get_page(pfn_to_page(pfn));
Anthony Liguori35149e22008-04-02 14:46:56 -05001283}
1284EXPORT_SYMBOL_GPL(kvm_get_pfn);
1285
Izik Eidus195aefd2007-10-01 22:14:18 +02001286static int next_segment(unsigned long len, int offset)
1287{
1288 if (len > PAGE_SIZE - offset)
1289 return PAGE_SIZE - offset;
1290 else
1291 return len;
1292}
1293
1294int kvm_read_guest_page(struct kvm *kvm, gfn_t gfn, void *data, int offset,
1295 int len)
1296{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001297 int r;
1298 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001299
Izik Eiduse0506bc2007-11-11 22:10:22 +02001300 addr = gfn_to_hva(kvm, gfn);
1301 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001302 return -EFAULT;
Takuya Yoshikawafa3d3152011-05-07 16:35:38 +09001303 r = __copy_from_user(data, (void __user *)addr + offset, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001304 if (r)
1305 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001306 return 0;
1307}
1308EXPORT_SYMBOL_GPL(kvm_read_guest_page);
1309
1310int kvm_read_guest(struct kvm *kvm, gpa_t gpa, void *data, unsigned long len)
1311{
1312 gfn_t gfn = gpa >> PAGE_SHIFT;
1313 int seg;
1314 int offset = offset_in_page(gpa);
1315 int ret;
1316
1317 while ((seg = next_segment(len, offset)) != 0) {
1318 ret = kvm_read_guest_page(kvm, gfn, data, offset, seg);
1319 if (ret < 0)
1320 return ret;
1321 offset = 0;
1322 len -= seg;
1323 data += seg;
1324 ++gfn;
1325 }
1326 return 0;
1327}
1328EXPORT_SYMBOL_GPL(kvm_read_guest);
1329
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001330int kvm_read_guest_atomic(struct kvm *kvm, gpa_t gpa, void *data,
1331 unsigned long len)
1332{
1333 int r;
1334 unsigned long addr;
1335 gfn_t gfn = gpa >> PAGE_SHIFT;
1336 int offset = offset_in_page(gpa);
1337
1338 addr = gfn_to_hva(kvm, gfn);
1339 if (kvm_is_error_hva(addr))
1340 return -EFAULT;
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001341 pagefault_disable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001342 r = __copy_from_user_inatomic(data, (void __user *)addr + offset, len);
Andrea Arcangeli0aac03f2008-01-30 19:57:35 +01001343 pagefault_enable();
Marcelo Tosatti7ec54582007-12-20 19:18:23 -05001344 if (r)
1345 return -EFAULT;
1346 return 0;
1347}
1348EXPORT_SYMBOL(kvm_read_guest_atomic);
1349
Izik Eidus195aefd2007-10-01 22:14:18 +02001350int kvm_write_guest_page(struct kvm *kvm, gfn_t gfn, const void *data,
1351 int offset, int len)
1352{
Izik Eiduse0506bc2007-11-11 22:10:22 +02001353 int r;
1354 unsigned long addr;
Izik Eidus195aefd2007-10-01 22:14:18 +02001355
Izik Eiduse0506bc2007-11-11 22:10:22 +02001356 addr = gfn_to_hva(kvm, gfn);
1357 if (kvm_is_error_hva(addr))
Izik Eidus195aefd2007-10-01 22:14:18 +02001358 return -EFAULT;
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001359 r = __copy_to_user((void __user *)addr + offset, data, len);
Izik Eiduse0506bc2007-11-11 22:10:22 +02001360 if (r)
1361 return -EFAULT;
Izik Eidus195aefd2007-10-01 22:14:18 +02001362 mark_page_dirty(kvm, gfn);
1363 return 0;
1364}
1365EXPORT_SYMBOL_GPL(kvm_write_guest_page);
1366
1367int kvm_write_guest(struct kvm *kvm, gpa_t gpa, const void *data,
1368 unsigned long len)
1369{
1370 gfn_t gfn = gpa >> PAGE_SHIFT;
1371 int seg;
1372 int offset = offset_in_page(gpa);
1373 int ret;
1374
1375 while ((seg = next_segment(len, offset)) != 0) {
1376 ret = kvm_write_guest_page(kvm, gfn, data, offset, seg);
1377 if (ret < 0)
1378 return ret;
1379 offset = 0;
1380 len -= seg;
1381 data += seg;
1382 ++gfn;
1383 }
1384 return 0;
1385}
1386
Gleb Natapov49c77542010-10-18 15:22:23 +02001387int kvm_gfn_to_hva_cache_init(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1388 gpa_t gpa)
1389{
1390 struct kvm_memslots *slots = kvm_memslots(kvm);
1391 int offset = offset_in_page(gpa);
1392 gfn_t gfn = gpa >> PAGE_SHIFT;
1393
1394 ghc->gpa = gpa;
1395 ghc->generation = slots->generation;
Paul Mackerras9d4cba72012-01-12 20:09:51 +00001396 ghc->memslot = gfn_to_memslot(kvm, gfn);
Gleb Natapov49c77542010-10-18 15:22:23 +02001397 ghc->hva = gfn_to_hva_many(ghc->memslot, gfn, NULL);
1398 if (!kvm_is_error_hva(ghc->hva))
1399 ghc->hva += offset;
1400 else
1401 return -EFAULT;
1402
1403 return 0;
1404}
1405EXPORT_SYMBOL_GPL(kvm_gfn_to_hva_cache_init);
1406
1407int kvm_write_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1408 void *data, unsigned long len)
1409{
1410 struct kvm_memslots *slots = kvm_memslots(kvm);
1411 int r;
1412
1413 if (slots->generation != ghc->generation)
1414 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1415
1416 if (kvm_is_error_hva(ghc->hva))
1417 return -EFAULT;
1418
Xiao Guangrong8b0cedf2011-05-15 23:22:04 +08001419 r = __copy_to_user((void __user *)ghc->hva, data, len);
Gleb Natapov49c77542010-10-18 15:22:23 +02001420 if (r)
1421 return -EFAULT;
1422 mark_page_dirty_in_slot(kvm, ghc->memslot, ghc->gpa >> PAGE_SHIFT);
1423
1424 return 0;
1425}
1426EXPORT_SYMBOL_GPL(kvm_write_guest_cached);
1427
Gleb Natapove03b6442011-07-11 15:28:11 -04001428int kvm_read_guest_cached(struct kvm *kvm, struct gfn_to_hva_cache *ghc,
1429 void *data, unsigned long len)
1430{
1431 struct kvm_memslots *slots = kvm_memslots(kvm);
1432 int r;
1433
1434 if (slots->generation != ghc->generation)
1435 kvm_gfn_to_hva_cache_init(kvm, ghc, ghc->gpa);
1436
1437 if (kvm_is_error_hva(ghc->hva))
1438 return -EFAULT;
1439
1440 r = __copy_from_user(data, (void __user *)ghc->hva, len);
1441 if (r)
1442 return -EFAULT;
1443
1444 return 0;
1445}
1446EXPORT_SYMBOL_GPL(kvm_read_guest_cached);
1447
Izik Eidus195aefd2007-10-01 22:14:18 +02001448int kvm_clear_guest_page(struct kvm *kvm, gfn_t gfn, int offset, int len)
1449{
Heiko Carstens3bcc8a82010-10-27 17:21:21 +02001450 return kvm_write_guest_page(kvm, gfn, (const void *) empty_zero_page,
1451 offset, len);
Izik Eidus195aefd2007-10-01 22:14:18 +02001452}
1453EXPORT_SYMBOL_GPL(kvm_clear_guest_page);
1454
1455int kvm_clear_guest(struct kvm *kvm, gpa_t gpa, unsigned long len)
1456{
1457 gfn_t gfn = gpa >> PAGE_SHIFT;
1458 int seg;
1459 int offset = offset_in_page(gpa);
1460 int ret;
1461
1462 while ((seg = next_segment(len, offset)) != 0) {
1463 ret = kvm_clear_guest_page(kvm, gfn, offset, seg);
1464 if (ret < 0)
1465 return ret;
1466 offset = 0;
1467 len -= seg;
1468 ++gfn;
1469 }
1470 return 0;
1471}
1472EXPORT_SYMBOL_GPL(kvm_clear_guest);
1473
Gleb Natapov49c77542010-10-18 15:22:23 +02001474void mark_page_dirty_in_slot(struct kvm *kvm, struct kvm_memory_slot *memslot,
1475 gfn_t gfn)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001476{
Rusty Russell7e9d6192007-07-31 20:41:14 +10001477 if (memslot && memslot->dirty_bitmap) {
1478 unsigned long rel_gfn = gfn - memslot->base_gfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001479
Takuya Yoshikawa50e92b32012-01-04 15:06:43 +09001480 if (!test_and_set_bit_le(rel_gfn, memslot->dirty_bitmap))
Takuya Yoshikawa7850ac52011-11-14 18:23:34 +09001481 memslot->nr_dirty_pages++;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001482 }
1483}
1484
Gleb Natapov49c77542010-10-18 15:22:23 +02001485void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
1486{
1487 struct kvm_memory_slot *memslot;
1488
1489 memslot = gfn_to_memslot(kvm, gfn);
1490 mark_page_dirty_in_slot(kvm, memslot, gfn);
1491}
1492
Eddie Dongb6958ce2007-07-18 12:15:21 +03001493/*
1494 * The vCPU has executed a HLT instruction with in-kernel mode enabled.
1495 */
Hollis Blanchard8776e512007-10-31 17:24:24 -05001496void kvm_vcpu_block(struct kvm_vcpu *vcpu)
Eddie Dongb6958ce2007-07-18 12:15:21 +03001497{
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001498 DEFINE_WAIT(wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001499
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001500 for (;;) {
1501 prepare_to_wait(&vcpu->wq, &wait, TASK_INTERRUPTIBLE);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001502
Gleb Natapova1b37102009-07-09 15:33:52 +03001503 if (kvm_arch_vcpu_runnable(vcpu)) {
Avi Kivitya8eeb042010-05-10 12:34:53 +03001504 kvm_make_request(KVM_REQ_UNHALT, vcpu);
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001505 break;
Marcelo Tosattid7690172008-09-08 15:23:48 -03001506 }
Gleb Natapov09cec752009-03-23 15:11:44 +02001507 if (kvm_cpu_has_pending_timer(vcpu))
1508 break;
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001509 if (signal_pending(current))
1510 break;
1511
Eddie Dongb6958ce2007-07-18 12:15:21 +03001512 schedule();
Eddie Dongb6958ce2007-07-18 12:15:21 +03001513 }
1514
Marcelo Tosattie5c239c2008-05-08 19:47:01 -03001515 finish_wait(&vcpu->wq, &wait);
Eddie Dongb6958ce2007-07-18 12:15:21 +03001516}
1517
Avi Kivity6aa8b732006-12-10 02:21:36 -08001518void kvm_resched(struct kvm_vcpu *vcpu)
1519{
Yaozu Dong3fca0362007-04-25 16:49:19 +03001520 if (!need_resched())
1521 return;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001522 cond_resched();
Avi Kivity6aa8b732006-12-10 02:21:36 -08001523}
1524EXPORT_SYMBOL_GPL(kvm_resched);
1525
Rik van Riel217ece62011-02-01 09:53:28 -05001526void kvm_vcpu_on_spin(struct kvm_vcpu *me)
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001527{
Rik van Riel217ece62011-02-01 09:53:28 -05001528 struct kvm *kvm = me->kvm;
1529 struct kvm_vcpu *vcpu;
1530 int last_boosted_vcpu = me->kvm->last_boosted_vcpu;
1531 int yielded = 0;
1532 int pass;
1533 int i;
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001534
Rik van Riel217ece62011-02-01 09:53:28 -05001535 /*
1536 * We boost the priority of a VCPU that is runnable but not
1537 * currently running, because it got preempted by something
1538 * else and called schedule in __vcpu_run. Hopefully that
1539 * VCPU is holding the lock that we need and will release it.
1540 * We approximate round-robin by starting at the last boosted VCPU.
1541 */
1542 for (pass = 0; pass < 2 && !yielded; pass++) {
1543 kvm_for_each_vcpu(i, vcpu, kvm) {
1544 struct task_struct *task = NULL;
1545 struct pid *pid;
1546 if (!pass && i < last_boosted_vcpu) {
1547 i = last_boosted_vcpu;
1548 continue;
1549 } else if (pass && i > last_boosted_vcpu)
1550 break;
1551 if (vcpu == me)
1552 continue;
1553 if (waitqueue_active(&vcpu->wq))
1554 continue;
1555 rcu_read_lock();
1556 pid = rcu_dereference(vcpu->pid);
1557 if (pid)
1558 task = get_pid_task(vcpu->pid, PIDTYPE_PID);
1559 rcu_read_unlock();
1560 if (!task)
1561 continue;
1562 if (task->flags & PF_VCPU) {
1563 put_task_struct(task);
1564 continue;
1565 }
1566 if (yield_to(task, 1)) {
1567 put_task_struct(task);
1568 kvm->last_boosted_vcpu = i;
1569 yielded = 1;
1570 break;
1571 }
1572 put_task_struct(task);
1573 }
1574 }
Zhai, Edwind255f4f2009-10-09 18:03:20 +08001575}
1576EXPORT_SYMBOL_GPL(kvm_vcpu_on_spin);
1577
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001578static int kvm_vcpu_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001579{
1580 struct kvm_vcpu *vcpu = vma->vm_file->private_data;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001581 struct page *page;
1582
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001583 if (vmf->pgoff == 0)
Avi Kivity039576c2007-03-20 12:46:50 +02001584 page = virt_to_page(vcpu->run);
Avi Kivity09566762008-01-23 18:14:23 +02001585#ifdef CONFIG_X86
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001586 else if (vmf->pgoff == KVM_PIO_PAGE_OFFSET)
Zhang Xiantaoad312c72007-12-13 23:50:52 +08001587 page = virt_to_page(vcpu->arch.pio_data);
Avi Kivity09566762008-01-23 18:14:23 +02001588#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02001589#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1590 else if (vmf->pgoff == KVM_COALESCED_MMIO_PAGE_OFFSET)
1591 page = virt_to_page(vcpu->kvm->coalesced_mmio_ring);
1592#endif
Avi Kivity039576c2007-03-20 12:46:50 +02001593 else
Carsten Otte5b1c1492012-01-04 10:25:23 +01001594 return kvm_arch_vcpu_fault(vcpu, vmf);
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001595 get_page(page);
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001596 vmf->page = page;
1597 return 0;
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001598}
1599
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04001600static const struct vm_operations_struct kvm_vcpu_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11001601 .fault = kvm_vcpu_fault,
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001602};
1603
1604static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
1605{
1606 vma->vm_ops = &kvm_vcpu_vm_ops;
1607 return 0;
1608}
1609
Avi Kivitybccf2152007-02-21 18:04:26 +02001610static int kvm_vcpu_release(struct inode *inode, struct file *filp)
1611{
1612 struct kvm_vcpu *vcpu = filp->private_data;
1613
Al Viro66c0b392008-04-19 20:33:56 +01001614 kvm_put_kvm(vcpu->kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001615 return 0;
1616}
1617
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01001618static struct file_operations kvm_vcpu_fops = {
Avi Kivitybccf2152007-02-21 18:04:26 +02001619 .release = kvm_vcpu_release,
1620 .unlocked_ioctl = kvm_vcpu_ioctl,
Alexander Graf1dda6062011-06-08 02:45:37 +02001621#ifdef CONFIG_COMPAT
1622 .compat_ioctl = kvm_vcpu_compat_ioctl,
1623#endif
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001624 .mmap = kvm_vcpu_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02001625 .llseek = noop_llseek,
Avi Kivitybccf2152007-02-21 18:04:26 +02001626};
1627
1628/*
1629 * Allocates an inode for the vcpu.
1630 */
1631static int create_vcpu_fd(struct kvm_vcpu *vcpu)
1632{
Roland Dreier628ff7c2009-12-18 09:41:24 -08001633 return anon_inode_getfd("kvm-vcpu", &kvm_vcpu_fops, vcpu, O_RDWR);
Avi Kivitybccf2152007-02-21 18:04:26 +02001634}
1635
Avi Kivityc5ea7662007-02-20 18:41:05 +02001636/*
1637 * Creates some virtual cpus. Good luck creating more than one.
1638 */
Gleb Natapov73880c82009-06-09 15:56:28 +03001639static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, u32 id)
Avi Kivityc5ea7662007-02-20 18:41:05 +02001640{
1641 int r;
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001642 struct kvm_vcpu *vcpu, *v;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001643
Gleb Natapov73880c82009-06-09 15:56:28 +03001644 vcpu = kvm_arch_vcpu_create(kvm, id);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001645 if (IS_ERR(vcpu))
1646 return PTR_ERR(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001647
Avi Kivity15ad7142007-07-11 18:17:21 +03001648 preempt_notifier_init(&vcpu->preempt_notifier, &kvm_preempt_ops);
1649
Avi Kivity26e52152007-11-20 15:30:24 +02001650 r = kvm_arch_vcpu_setup(vcpu);
1651 if (r)
Jan Kiszkad7805922011-05-23 10:33:05 +02001652 goto vcpu_destroy;
Avi Kivity26e52152007-11-20 15:30:24 +02001653
Shaohua Li11ec2802007-07-23 14:51:37 +08001654 mutex_lock(&kvm->lock);
Avi Kivity3e515702012-03-05 14:23:29 +02001655 if (!kvm_vcpu_compatible(vcpu)) {
1656 r = -EINVAL;
1657 goto unlock_vcpu_destroy;
1658 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001659 if (atomic_read(&kvm->online_vcpus) == KVM_MAX_VCPUS) {
1660 r = -EINVAL;
Jan Kiszkad7805922011-05-23 10:33:05 +02001661 goto unlock_vcpu_destroy;
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001662 }
Gleb Natapov73880c82009-06-09 15:56:28 +03001663
Gleb Natapov988a2ca2009-06-09 15:56:29 +03001664 kvm_for_each_vcpu(r, v, kvm)
1665 if (v->vcpu_id == id) {
Gleb Natapov73880c82009-06-09 15:56:28 +03001666 r = -EEXIST;
Jan Kiszkad7805922011-05-23 10:33:05 +02001667 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001668 }
1669
1670 BUG_ON(kvm->vcpus[atomic_read(&kvm->online_vcpus)]);
Rusty Russellfb3f0f52007-07-27 17:16:56 +10001671
1672 /* Now it's all set up, let userspace reach it */
Al Viro66c0b392008-04-19 20:33:56 +01001673 kvm_get_kvm(kvm);
Avi Kivitybccf2152007-02-21 18:04:26 +02001674 r = create_vcpu_fd(vcpu);
Gleb Natapov73880c82009-06-09 15:56:28 +03001675 if (r < 0) {
1676 kvm_put_kvm(kvm);
Jan Kiszkad7805922011-05-23 10:33:05 +02001677 goto unlock_vcpu_destroy;
Gleb Natapov73880c82009-06-09 15:56:28 +03001678 }
1679
1680 kvm->vcpus[atomic_read(&kvm->online_vcpus)] = vcpu;
1681 smp_wmb();
1682 atomic_inc(&kvm->online_vcpus);
1683
Gleb Natapov73880c82009-06-09 15:56:28 +03001684 mutex_unlock(&kvm->lock);
Avi Kivitybccf2152007-02-21 18:04:26 +02001685 return r;
Avi Kivityc5ea7662007-02-20 18:41:05 +02001686
Jan Kiszkad7805922011-05-23 10:33:05 +02001687unlock_vcpu_destroy:
Glauber Costa7d8fece2008-09-17 23:16:59 -03001688 mutex_unlock(&kvm->lock);
Jan Kiszkad7805922011-05-23 10:33:05 +02001689vcpu_destroy:
Hollis Blanchardd40ccc62007-11-19 14:04:43 -06001690 kvm_arch_vcpu_destroy(vcpu);
Avi Kivityc5ea7662007-02-20 18:41:05 +02001691 return r;
1692}
1693
Avi Kivity1961d272007-03-05 19:46:05 +02001694static int kvm_vcpu_ioctl_set_sigmask(struct kvm_vcpu *vcpu, sigset_t *sigset)
1695{
1696 if (sigset) {
1697 sigdelsetmask(sigset, sigmask(SIGKILL)|sigmask(SIGSTOP));
1698 vcpu->sigset_active = 1;
1699 vcpu->sigset = *sigset;
1700 } else
1701 vcpu->sigset_active = 0;
1702 return 0;
1703}
1704
Avi Kivitybccf2152007-02-21 18:04:26 +02001705static long kvm_vcpu_ioctl(struct file *filp,
1706 unsigned int ioctl, unsigned long arg)
Avi Kivity6aa8b732006-12-10 02:21:36 -08001707{
Avi Kivitybccf2152007-02-21 18:04:26 +02001708 struct kvm_vcpu *vcpu = filp->private_data;
Al Viro2f3669872007-02-09 16:38:35 +00001709 void __user *argp = (void __user *)arg;
Carsten Otte313a3dc2007-10-11 19:16:52 +02001710 int r;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001711 struct kvm_fpu *fpu = NULL;
1712 struct kvm_sregs *kvm_sregs = NULL;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001713
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001714 if (vcpu->kvm->mm != current->mm)
1715 return -EIO;
Avi Kivity2122ff52010-05-13 11:25:04 +03001716
1717#if defined(CONFIG_S390) || defined(CONFIG_PPC)
1718 /*
1719 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
1720 * so vcpu_load() would break it.
1721 */
1722 if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_INTERRUPT)
1723 return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
1724#endif
1725
1726
1727 vcpu_load(vcpu);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001728 switch (ioctl) {
Avi Kivity9a2bb7f2007-02-22 12:58:31 +02001729 case KVM_RUN:
Avi Kivityf0fe5102007-03-07 13:11:17 +02001730 r = -EINVAL;
1731 if (arg)
1732 goto out;
Hollis Blanchardb6c7a5d2007-11-01 14:16:10 -05001733 r = kvm_arch_vcpu_ioctl_run(vcpu, vcpu->run);
Gleb Natapov64be5002010-10-24 16:49:08 +02001734 trace_kvm_userspace_exit(vcpu->run->exit_reason, r);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001735 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001736 case KVM_GET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001737 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001738
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001739 r = -ENOMEM;
1740 kvm_regs = kzalloc(sizeof(struct kvm_regs), GFP_KERNEL);
1741 if (!kvm_regs)
1742 goto out;
1743 r = kvm_arch_vcpu_ioctl_get_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001744 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001745 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001746 r = -EFAULT;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001747 if (copy_to_user(argp, kvm_regs, sizeof(struct kvm_regs)))
1748 goto out_free1;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001749 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001750out_free1:
1751 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001752 break;
1753 }
1754 case KVM_SET_REGS: {
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001755 struct kvm_regs *kvm_regs;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001756
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001757 r = -ENOMEM;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001758 kvm_regs = memdup_user(argp, sizeof(*kvm_regs));
1759 if (IS_ERR(kvm_regs)) {
1760 r = PTR_ERR(kvm_regs);
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001761 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001762 }
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001763 r = kvm_arch_vcpu_ioctl_set_regs(vcpu, kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001764 if (r)
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001765 goto out_free2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001766 r = 0;
Xiantao Zhang3e4bb3a2008-02-25 18:52:20 +08001767out_free2:
1768 kfree(kvm_regs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001769 break;
1770 }
1771 case KVM_GET_SREGS: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001772 kvm_sregs = kzalloc(sizeof(struct kvm_sregs), GFP_KERNEL);
1773 r = -ENOMEM;
1774 if (!kvm_sregs)
1775 goto out;
1776 r = kvm_arch_vcpu_ioctl_get_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001777 if (r)
1778 goto out;
1779 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001780 if (copy_to_user(argp, kvm_sregs, sizeof(struct kvm_sregs)))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001781 goto out;
1782 r = 0;
1783 break;
1784 }
1785 case KVM_SET_SREGS: {
Sasha Levinff5c2c02011-12-04 19:36:29 +02001786 kvm_sregs = memdup_user(argp, sizeof(*kvm_sregs));
1787 if (IS_ERR(kvm_sregs)) {
1788 r = PTR_ERR(kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001789 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001790 }
Dave Hansenfa3795a2008-08-11 10:01:46 -07001791 r = kvm_arch_vcpu_ioctl_set_sregs(vcpu, kvm_sregs);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001792 if (r)
1793 goto out;
1794 r = 0;
1795 break;
1796 }
Marcelo Tosatti62d9f0d2008-04-11 13:24:45 -03001797 case KVM_GET_MP_STATE: {
1798 struct kvm_mp_state mp_state;
1799
1800 r = kvm_arch_vcpu_ioctl_get_mpstate(vcpu, &mp_state);
1801 if (r)
1802 goto out;
1803 r = -EFAULT;
1804 if (copy_to_user(argp, &mp_state, sizeof mp_state))
1805 goto out;
1806 r = 0;
1807 break;
1808 }
1809 case KVM_SET_MP_STATE: {
1810 struct kvm_mp_state mp_state;
1811
1812 r = -EFAULT;
1813 if (copy_from_user(&mp_state, argp, sizeof mp_state))
1814 goto out;
1815 r = kvm_arch_vcpu_ioctl_set_mpstate(vcpu, &mp_state);
1816 if (r)
1817 goto out;
1818 r = 0;
1819 break;
1820 }
Avi Kivity6aa8b732006-12-10 02:21:36 -08001821 case KVM_TRANSLATE: {
1822 struct kvm_translation tr;
1823
1824 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001825 if (copy_from_user(&tr, argp, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001826 goto out;
Zhang Xiantao8b006792007-11-16 13:05:55 +08001827 r = kvm_arch_vcpu_ioctl_translate(vcpu, &tr);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001828 if (r)
1829 goto out;
1830 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001831 if (copy_to_user(argp, &tr, sizeof tr))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001832 goto out;
1833 r = 0;
1834 break;
1835 }
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001836 case KVM_SET_GUEST_DEBUG: {
1837 struct kvm_guest_debug dbg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08001838
1839 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001840 if (copy_from_user(&dbg, argp, sizeof dbg))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001841 goto out;
Jan Kiszkad0bfb942008-12-15 13:52:10 +01001842 r = kvm_arch_vcpu_ioctl_set_guest_debug(vcpu, &dbg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001843 if (r)
1844 goto out;
1845 r = 0;
1846 break;
1847 }
Avi Kivity1961d272007-03-05 19:46:05 +02001848 case KVM_SET_SIGNAL_MASK: {
1849 struct kvm_signal_mask __user *sigmask_arg = argp;
1850 struct kvm_signal_mask kvm_sigmask;
1851 sigset_t sigset, *p;
1852
1853 p = NULL;
1854 if (argp) {
1855 r = -EFAULT;
1856 if (copy_from_user(&kvm_sigmask, argp,
1857 sizeof kvm_sigmask))
1858 goto out;
1859 r = -EINVAL;
1860 if (kvm_sigmask.len != sizeof sigset)
1861 goto out;
1862 r = -EFAULT;
1863 if (copy_from_user(&sigset, sigmask_arg->sigset,
1864 sizeof sigset))
1865 goto out;
1866 p = &sigset;
1867 }
Andi Kleen376d41f2010-06-10 13:10:47 +02001868 r = kvm_vcpu_ioctl_set_sigmask(vcpu, p);
Avi Kivity1961d272007-03-05 19:46:05 +02001869 break;
1870 }
Avi Kivityb8836732007-04-01 16:34:31 +03001871 case KVM_GET_FPU: {
Dave Hansenfa3795a2008-08-11 10:01:46 -07001872 fpu = kzalloc(sizeof(struct kvm_fpu), GFP_KERNEL);
1873 r = -ENOMEM;
1874 if (!fpu)
1875 goto out;
1876 r = kvm_arch_vcpu_ioctl_get_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001877 if (r)
1878 goto out;
1879 r = -EFAULT;
Dave Hansenfa3795a2008-08-11 10:01:46 -07001880 if (copy_to_user(argp, fpu, sizeof(struct kvm_fpu)))
Avi Kivityb8836732007-04-01 16:34:31 +03001881 goto out;
1882 r = 0;
1883 break;
1884 }
1885 case KVM_SET_FPU: {
Sasha Levinff5c2c02011-12-04 19:36:29 +02001886 fpu = memdup_user(argp, sizeof(*fpu));
1887 if (IS_ERR(fpu)) {
1888 r = PTR_ERR(fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001889 goto out;
Sasha Levinff5c2c02011-12-04 19:36:29 +02001890 }
Dave Hansenfa3795a2008-08-11 10:01:46 -07001891 r = kvm_arch_vcpu_ioctl_set_fpu(vcpu, fpu);
Avi Kivityb8836732007-04-01 16:34:31 +03001892 if (r)
1893 goto out;
1894 r = 0;
1895 break;
1896 }
Avi Kivitybccf2152007-02-21 18:04:26 +02001897 default:
Carsten Otte313a3dc2007-10-11 19:16:52 +02001898 r = kvm_arch_vcpu_ioctl(filp, ioctl, arg);
Avi Kivitybccf2152007-02-21 18:04:26 +02001899 }
1900out:
Avi Kivity2122ff52010-05-13 11:25:04 +03001901 vcpu_put(vcpu);
Dave Hansenfa3795a2008-08-11 10:01:46 -07001902 kfree(fpu);
1903 kfree(kvm_sregs);
Avi Kivitybccf2152007-02-21 18:04:26 +02001904 return r;
1905}
1906
Alexander Graf1dda6062011-06-08 02:45:37 +02001907#ifdef CONFIG_COMPAT
1908static long kvm_vcpu_compat_ioctl(struct file *filp,
1909 unsigned int ioctl, unsigned long arg)
1910{
1911 struct kvm_vcpu *vcpu = filp->private_data;
1912 void __user *argp = compat_ptr(arg);
1913 int r;
1914
1915 if (vcpu->kvm->mm != current->mm)
1916 return -EIO;
1917
1918 switch (ioctl) {
1919 case KVM_SET_SIGNAL_MASK: {
1920 struct kvm_signal_mask __user *sigmask_arg = argp;
1921 struct kvm_signal_mask kvm_sigmask;
1922 compat_sigset_t csigset;
1923 sigset_t sigset;
1924
1925 if (argp) {
1926 r = -EFAULT;
1927 if (copy_from_user(&kvm_sigmask, argp,
1928 sizeof kvm_sigmask))
1929 goto out;
1930 r = -EINVAL;
1931 if (kvm_sigmask.len != sizeof csigset)
1932 goto out;
1933 r = -EFAULT;
1934 if (copy_from_user(&csigset, sigmask_arg->sigset,
1935 sizeof csigset))
1936 goto out;
1937 }
1938 sigset_from_compat(&sigset, &csigset);
1939 r = kvm_vcpu_ioctl_set_sigmask(vcpu, &sigset);
1940 break;
1941 }
1942 default:
1943 r = kvm_vcpu_ioctl(filp, ioctl, arg);
1944 }
1945
1946out:
1947 return r;
1948}
1949#endif
1950
Avi Kivitybccf2152007-02-21 18:04:26 +02001951static long kvm_vm_ioctl(struct file *filp,
1952 unsigned int ioctl, unsigned long arg)
1953{
1954 struct kvm *kvm = filp->private_data;
1955 void __user *argp = (void __user *)arg;
Carsten Otte1fe779f2007-10-29 16:08:35 +01001956 int r;
Avi Kivitybccf2152007-02-21 18:04:26 +02001957
Avi Kivity6d4e4c42007-11-21 16:41:05 +02001958 if (kvm->mm != current->mm)
1959 return -EIO;
Avi Kivitybccf2152007-02-21 18:04:26 +02001960 switch (ioctl) {
1961 case KVM_CREATE_VCPU:
1962 r = kvm_vm_ioctl_create_vcpu(kvm, arg);
1963 if (r < 0)
1964 goto out;
1965 break;
Izik Eidus6fc138d2007-10-09 19:20:39 +02001966 case KVM_SET_USER_MEMORY_REGION: {
1967 struct kvm_userspace_memory_region kvm_userspace_mem;
1968
1969 r = -EFAULT;
1970 if (copy_from_user(&kvm_userspace_mem, argp,
1971 sizeof kvm_userspace_mem))
1972 goto out;
1973
1974 r = kvm_vm_ioctl_set_memory_region(kvm, &kvm_userspace_mem, 1);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001975 if (r)
1976 goto out;
1977 break;
1978 }
1979 case KVM_GET_DIRTY_LOG: {
1980 struct kvm_dirty_log log;
1981
1982 r = -EFAULT;
Al Viro2f3669872007-02-09 16:38:35 +00001983 if (copy_from_user(&log, argp, sizeof log))
Avi Kivity6aa8b732006-12-10 02:21:36 -08001984 goto out;
Avi Kivity2c6f5df2007-02-20 18:27:58 +02001985 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
Avi Kivity6aa8b732006-12-10 02:21:36 -08001986 if (r)
1987 goto out;
1988 break;
1989 }
Laurent Vivier5f94c172008-05-30 16:05:54 +02001990#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
1991 case KVM_REGISTER_COALESCED_MMIO: {
1992 struct kvm_coalesced_mmio_zone zone;
1993 r = -EFAULT;
1994 if (copy_from_user(&zone, argp, sizeof zone))
1995 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02001996 r = kvm_vm_ioctl_register_coalesced_mmio(kvm, &zone);
1997 if (r)
1998 goto out;
1999 r = 0;
2000 break;
2001 }
2002 case KVM_UNREGISTER_COALESCED_MMIO: {
2003 struct kvm_coalesced_mmio_zone zone;
2004 r = -EFAULT;
2005 if (copy_from_user(&zone, argp, sizeof zone))
2006 goto out;
Laurent Vivier5f94c172008-05-30 16:05:54 +02002007 r = kvm_vm_ioctl_unregister_coalesced_mmio(kvm, &zone);
2008 if (r)
2009 goto out;
2010 r = 0;
2011 break;
2012 }
2013#endif
Gregory Haskins721eecb2009-05-20 10:30:49 -04002014 case KVM_IRQFD: {
2015 struct kvm_irqfd data;
2016
2017 r = -EFAULT;
2018 if (copy_from_user(&data, argp, sizeof data))
2019 goto out;
2020 r = kvm_irqfd(kvm, data.fd, data.gsi, data.flags);
2021 break;
2022 }
Gregory Haskinsd34e6b12009-07-07 17:08:49 -04002023 case KVM_IOEVENTFD: {
2024 struct kvm_ioeventfd data;
2025
2026 r = -EFAULT;
2027 if (copy_from_user(&data, argp, sizeof data))
2028 goto out;
2029 r = kvm_ioeventfd(kvm, &data);
2030 break;
2031 }
Gleb Natapov73880c82009-06-09 15:56:28 +03002032#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2033 case KVM_SET_BOOT_CPU_ID:
2034 r = 0;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002035 mutex_lock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002036 if (atomic_read(&kvm->online_vcpus) != 0)
2037 r = -EBUSY;
2038 else
2039 kvm->bsp_vcpu_id = arg;
Marcelo Tosatti894a9c52009-06-23 15:05:14 -03002040 mutex_unlock(&kvm->lock);
Gleb Natapov73880c82009-06-09 15:56:28 +03002041 break;
2042#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002043 default:
Carsten Otte1fe779f2007-10-29 16:08:35 +01002044 r = kvm_arch_vm_ioctl(filp, ioctl, arg);
Avi Kivitybfd99ff2009-08-26 14:57:50 +03002045 if (r == -ENOTTY)
2046 r = kvm_vm_ioctl_assigned_device(kvm, ioctl, arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002047 }
2048out:
2049 return r;
2050}
2051
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002052#ifdef CONFIG_COMPAT
2053struct compat_kvm_dirty_log {
2054 __u32 slot;
2055 __u32 padding1;
2056 union {
2057 compat_uptr_t dirty_bitmap; /* one bit per page */
2058 __u64 padding2;
2059 };
2060};
2061
2062static long kvm_vm_compat_ioctl(struct file *filp,
2063 unsigned int ioctl, unsigned long arg)
2064{
2065 struct kvm *kvm = filp->private_data;
2066 int r;
2067
2068 if (kvm->mm != current->mm)
2069 return -EIO;
2070 switch (ioctl) {
2071 case KVM_GET_DIRTY_LOG: {
2072 struct compat_kvm_dirty_log compat_log;
2073 struct kvm_dirty_log log;
2074
2075 r = -EFAULT;
2076 if (copy_from_user(&compat_log, (void __user *)arg,
2077 sizeof(compat_log)))
2078 goto out;
2079 log.slot = compat_log.slot;
2080 log.padding1 = compat_log.padding1;
2081 log.padding2 = compat_log.padding2;
2082 log.dirty_bitmap = compat_ptr(compat_log.dirty_bitmap);
2083
2084 r = kvm_vm_ioctl_get_dirty_log(kvm, &log);
2085 if (r)
2086 goto out;
2087 break;
2088 }
2089 default:
2090 r = kvm_vm_ioctl(filp, ioctl, arg);
2091 }
2092
2093out:
2094 return r;
2095}
2096#endif
2097
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002098static int kvm_vm_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
Avi Kivityf17abe92007-02-21 19:28:04 +02002099{
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002100 struct page *page[1];
2101 unsigned long addr;
2102 int npages;
2103 gfn_t gfn = vmf->pgoff;
Avi Kivityf17abe92007-02-21 19:28:04 +02002104 struct kvm *kvm = vma->vm_file->private_data;
Avi Kivityf17abe92007-02-21 19:28:04 +02002105
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002106 addr = gfn_to_hva(kvm, gfn);
2107 if (kvm_is_error_hva(addr))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002108 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002109
2110 npages = get_user_pages(current, current->mm, addr, 1, 1, 0, page,
2111 NULL);
2112 if (unlikely(npages != 1))
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002113 return VM_FAULT_SIGBUS;
Marcelo Tosatti777b3f42008-09-16 20:54:46 -03002114
2115 vmf->page = page[0];
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002116 return 0;
Avi Kivityf17abe92007-02-21 19:28:04 +02002117}
2118
Alexey Dobriyanf0f37e22009-09-27 22:29:37 +04002119static const struct vm_operations_struct kvm_vm_vm_ops = {
npiggin@suse.dee4a533a2007-12-05 18:15:52 +11002120 .fault = kvm_vm_fault,
Avi Kivityf17abe92007-02-21 19:28:04 +02002121};
2122
2123static int kvm_vm_mmap(struct file *file, struct vm_area_struct *vma)
2124{
2125 vma->vm_ops = &kvm_vm_vm_ops;
2126 return 0;
2127}
2128
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002129static struct file_operations kvm_vm_fops = {
Avi Kivityf17abe92007-02-21 19:28:04 +02002130 .release = kvm_vm_release,
2131 .unlocked_ioctl = kvm_vm_ioctl,
Arnd Bergmann6ff58942009-10-22 14:19:27 +02002132#ifdef CONFIG_COMPAT
2133 .compat_ioctl = kvm_vm_compat_ioctl,
2134#endif
Avi Kivityf17abe92007-02-21 19:28:04 +02002135 .mmap = kvm_vm_mmap,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002136 .llseek = noop_llseek,
Avi Kivityf17abe92007-02-21 19:28:04 +02002137};
2138
Carsten Ottee08b9632012-01-04 10:25:20 +01002139static int kvm_dev_ioctl_create_vm(unsigned long type)
Avi Kivityf17abe92007-02-21 19:28:04 +02002140{
Heiko Carstensaac87632010-10-27 17:22:10 +02002141 int r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002142 struct kvm *kvm;
2143
Carsten Ottee08b9632012-01-04 10:25:20 +01002144 kvm = kvm_create_vm(type);
Avi Kivityd6d28162007-06-28 08:38:16 -04002145 if (IS_ERR(kvm))
2146 return PTR_ERR(kvm);
Takuya Yoshikawa6ce5a092010-03-15 22:13:30 +09002147#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2148 r = kvm_coalesced_mmio_init(kvm);
2149 if (r < 0) {
2150 kvm_put_kvm(kvm);
2151 return r;
2152 }
2153#endif
Heiko Carstensaac87632010-10-27 17:22:10 +02002154 r = anon_inode_getfd("kvm-vm", &kvm_vm_fops, kvm, O_RDWR);
2155 if (r < 0)
Al Viro66c0b392008-04-19 20:33:56 +01002156 kvm_put_kvm(kvm);
Avi Kivityf17abe92007-02-21 19:28:04 +02002157
Heiko Carstensaac87632010-10-27 17:22:10 +02002158 return r;
Avi Kivityf17abe92007-02-21 19:28:04 +02002159}
2160
Avi Kivity1a811b62008-12-08 18:25:27 +02002161static long kvm_dev_ioctl_check_extension_generic(long arg)
2162{
2163 switch (arg) {
Avi Kivityca9edae2008-12-08 18:29:29 +02002164 case KVM_CAP_USER_MEMORY:
Avi Kivity1a811b62008-12-08 18:25:27 +02002165 case KVM_CAP_DESTROY_MEMORY_REGION_WORKS:
Jan Kiszka4cd481f2009-04-13 11:59:32 +02002166 case KVM_CAP_JOIN_MEMORY_REGIONS_WORKS:
Gleb Natapov73880c82009-06-09 15:56:28 +03002167#ifdef CONFIG_KVM_APIC_ARCHITECTURE
2168 case KVM_CAP_SET_BOOT_CPU_ID:
2169#endif
Avi Kivitya9c73992009-11-04 11:54:59 +02002170 case KVM_CAP_INTERNAL_ERROR_DATA:
Avi Kivity1a811b62008-12-08 18:25:27 +02002171 return 1;
Avi Kivity399ec802008-11-19 13:58:46 +02002172#ifdef CONFIG_HAVE_KVM_IRQCHIP
2173 case KVM_CAP_IRQ_ROUTING:
Sheng Yang36463142009-03-16 16:33:43 +08002174 return KVM_MAX_IRQ_ROUTES;
Avi Kivity399ec802008-11-19 13:58:46 +02002175#endif
Avi Kivity1a811b62008-12-08 18:25:27 +02002176 default:
2177 break;
2178 }
2179 return kvm_dev_ioctl_check_extension(arg);
2180}
2181
Avi Kivityf17abe92007-02-21 19:28:04 +02002182static long kvm_dev_ioctl(struct file *filp,
2183 unsigned int ioctl, unsigned long arg)
2184{
Avi Kivity07c45a32007-03-07 13:05:38 +02002185 long r = -EINVAL;
Avi Kivityf17abe92007-02-21 19:28:04 +02002186
2187 switch (ioctl) {
2188 case KVM_GET_API_VERSION:
Avi Kivityf0fe5102007-03-07 13:11:17 +02002189 r = -EINVAL;
2190 if (arg)
2191 goto out;
Avi Kivityf17abe92007-02-21 19:28:04 +02002192 r = KVM_API_VERSION;
2193 break;
2194 case KVM_CREATE_VM:
Carsten Ottee08b9632012-01-04 10:25:20 +01002195 r = kvm_dev_ioctl_create_vm(arg);
Avi Kivityf17abe92007-02-21 19:28:04 +02002196 break;
Zhang Xiantao018d00d2007-11-15 23:07:47 +08002197 case KVM_CHECK_EXTENSION:
Avi Kivity1a811b62008-12-08 18:25:27 +02002198 r = kvm_dev_ioctl_check_extension_generic(arg);
Avi Kivity5d308f42007-03-01 17:56:20 +02002199 break;
Avi Kivity07c45a32007-03-07 13:05:38 +02002200 case KVM_GET_VCPU_MMAP_SIZE:
2201 r = -EINVAL;
2202 if (arg)
2203 goto out;
Avi Kivityadb1ff42008-01-24 15:13:08 +02002204 r = PAGE_SIZE; /* struct kvm_run */
2205#ifdef CONFIG_X86
2206 r += PAGE_SIZE; /* pio data page */
2207#endif
Laurent Vivier5f94c172008-05-30 16:05:54 +02002208#ifdef KVM_COALESCED_MMIO_PAGE_OFFSET
2209 r += PAGE_SIZE; /* coalesced mmio ring page */
2210#endif
Avi Kivity07c45a32007-03-07 13:05:38 +02002211 break;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002212 case KVM_TRACE_ENABLE:
2213 case KVM_TRACE_PAUSE:
2214 case KVM_TRACE_DISABLE:
Marcelo Tosatti2023a292009-06-18 11:47:28 -03002215 r = -EOPNOTSUPP;
Feng(Eric) Liud4c9ff22008-04-10 08:47:53 -04002216 break;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002217 default:
Carsten Otte043405e2007-10-10 17:16:19 +02002218 return kvm_arch_dev_ioctl(filp, ioctl, arg);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002219 }
2220out:
2221 return r;
2222}
2223
Avi Kivity6aa8b732006-12-10 02:21:36 -08002224static struct file_operations kvm_chardev_ops = {
Avi Kivity6aa8b732006-12-10 02:21:36 -08002225 .unlocked_ioctl = kvm_dev_ioctl,
2226 .compat_ioctl = kvm_dev_ioctl,
Arnd Bergmann6038f372010-08-15 18:52:59 +02002227 .llseek = noop_llseek,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002228};
2229
2230static struct miscdevice kvm_dev = {
Avi Kivitybbe44322007-03-04 13:27:36 +02002231 KVM_MINOR,
Avi Kivity6aa8b732006-12-10 02:21:36 -08002232 "kvm",
2233 &kvm_chardev_ops,
2234};
2235
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002236static void hardware_enable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002237{
2238 int cpu = raw_smp_processor_id();
Alexander Graf10474ae2009-09-15 11:37:46 +02002239 int r;
Avi Kivity1b6c0162007-05-24 13:03:52 +03002240
Rusty Russell7f59f492008-12-07 21:25:45 +10302241 if (cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002242 return;
Alexander Graf10474ae2009-09-15 11:37:46 +02002243
Rusty Russell7f59f492008-12-07 21:25:45 +10302244 cpumask_set_cpu(cpu, cpus_hardware_enabled);
Alexander Graf10474ae2009-09-15 11:37:46 +02002245
2246 r = kvm_arch_hardware_enable(NULL);
2247
2248 if (r) {
2249 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
2250 atomic_inc(&hardware_enable_failed);
2251 printk(KERN_INFO "kvm: enabling virtualization on "
2252 "CPU%d failed\n", cpu);
2253 }
Avi Kivity1b6c0162007-05-24 13:03:52 +03002254}
2255
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002256static void hardware_enable(void *junk)
2257{
Jan Kiszkae935b832011-02-08 12:55:33 +01002258 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002259 hardware_enable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002260 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002261}
2262
2263static void hardware_disable_nolock(void *junk)
Avi Kivity1b6c0162007-05-24 13:03:52 +03002264{
2265 int cpu = raw_smp_processor_id();
2266
Rusty Russell7f59f492008-12-07 21:25:45 +10302267 if (!cpumask_test_cpu(cpu, cpus_hardware_enabled))
Avi Kivity1b6c0162007-05-24 13:03:52 +03002268 return;
Rusty Russell7f59f492008-12-07 21:25:45 +10302269 cpumask_clear_cpu(cpu, cpus_hardware_enabled);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002270 kvm_arch_hardware_disable(NULL);
Avi Kivity1b6c0162007-05-24 13:03:52 +03002271}
2272
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002273static void hardware_disable(void *junk)
2274{
Jan Kiszkae935b832011-02-08 12:55:33 +01002275 raw_spin_lock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002276 hardware_disable_nolock(junk);
Jan Kiszkae935b832011-02-08 12:55:33 +01002277 raw_spin_unlock(&kvm_lock);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002278}
2279
Alexander Graf10474ae2009-09-15 11:37:46 +02002280static void hardware_disable_all_nolock(void)
2281{
2282 BUG_ON(!kvm_usage_count);
2283
2284 kvm_usage_count--;
2285 if (!kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002286 on_each_cpu(hardware_disable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002287}
2288
2289static void hardware_disable_all(void)
2290{
Jan Kiszkae935b832011-02-08 12:55:33 +01002291 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002292 hardware_disable_all_nolock();
Jan Kiszkae935b832011-02-08 12:55:33 +01002293 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002294}
2295
2296static int hardware_enable_all(void)
2297{
2298 int r = 0;
2299
Jan Kiszkae935b832011-02-08 12:55:33 +01002300 raw_spin_lock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002301
2302 kvm_usage_count++;
2303 if (kvm_usage_count == 1) {
2304 atomic_set(&hardware_enable_failed, 0);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002305 on_each_cpu(hardware_enable_nolock, NULL, 1);
Alexander Graf10474ae2009-09-15 11:37:46 +02002306
2307 if (atomic_read(&hardware_enable_failed)) {
2308 hardware_disable_all_nolock();
2309 r = -EBUSY;
2310 }
2311 }
2312
Jan Kiszkae935b832011-02-08 12:55:33 +01002313 raw_spin_unlock(&kvm_lock);
Alexander Graf10474ae2009-09-15 11:37:46 +02002314
2315 return r;
2316}
2317
Avi Kivity774c47f2007-02-12 00:54:47 -08002318static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
2319 void *v)
2320{
2321 int cpu = (long)v;
2322
Alexander Graf10474ae2009-09-15 11:37:46 +02002323 if (!kvm_usage_count)
2324 return NOTIFY_OK;
2325
Avi Kivity1a6f4d72007-11-11 18:37:32 +02002326 val &= ~CPU_TASKS_FROZEN;
Avi Kivity774c47f2007-02-12 00:54:47 -08002327 switch (val) {
Avi Kivitycec9ad22007-05-24 13:11:41 +03002328 case CPU_DYING:
Avi Kivity6ec8a8562007-08-19 15:57:26 +03002329 printk(KERN_INFO "kvm: disabling virtualization on CPU%d\n",
2330 cpu);
2331 hardware_disable(NULL);
2332 break;
Zachary Amsdenda908f22010-08-19 22:07:27 -10002333 case CPU_STARTING:
Jeremy Katz43934a32007-02-19 14:37:46 +02002334 printk(KERN_INFO "kvm: enabling virtualization on CPU%d\n",
2335 cpu);
Zachary Amsdenda908f22010-08-19 22:07:27 -10002336 hardware_enable(NULL);
Avi Kivity774c47f2007-02-12 00:54:47 -08002337 break;
2338 }
2339 return NOTIFY_OK;
2340}
2341
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002342
Avi Kivityb7c41452010-12-02 17:52:50 +02002343asmlinkage void kvm_spurious_fault(void)
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002344{
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002345 /* Fault while not rebooting. We want the trace. */
2346 BUG();
2347}
Avi Kivityb7c41452010-12-02 17:52:50 +02002348EXPORT_SYMBOL_GPL(kvm_spurious_fault);
Avi Kivity4ecac3f2008-05-13 13:23:38 +03002349
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002350static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
Mike Dayd77c26f2007-10-08 09:02:08 -04002351 void *v)
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002352{
Sheng Yang8e1c1812009-04-29 11:09:04 +08002353 /*
2354 * Some (well, at least mine) BIOSes hang on reboot if
2355 * in vmx root mode.
2356 *
2357 * And Intel TXT required VMX off for all cpu when system shutdown.
2358 */
2359 printk(KERN_INFO "kvm: exiting hardware virtualization\n");
2360 kvm_rebooting = true;
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002361 on_each_cpu(hardware_disable_nolock, NULL, 1);
Rusty Russell9a2b85c2007-07-17 23:17:55 +10002362 return NOTIFY_OK;
2363}
2364
2365static struct notifier_block kvm_reboot_notifier = {
2366 .notifier_call = kvm_reboot,
2367 .priority = 0,
2368};
2369
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002370static void kvm_io_bus_destroy(struct kvm_io_bus *bus)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002371{
2372 int i;
2373
2374 for (i = 0; i < bus->dev_count; i++) {
Sasha Levin743eeb02011-07-27 16:00:48 +03002375 struct kvm_io_device *pos = bus->range[i].dev;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002376
2377 kvm_iodevice_destructor(pos);
2378 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002379 kfree(bus);
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002380}
2381
Sasha Levin743eeb02011-07-27 16:00:48 +03002382int kvm_io_bus_sort_cmp(const void *p1, const void *p2)
2383{
2384 const struct kvm_io_range *r1 = p1;
2385 const struct kvm_io_range *r2 = p2;
2386
2387 if (r1->addr < r2->addr)
2388 return -1;
2389 if (r1->addr + r1->len > r2->addr + r2->len)
2390 return 1;
2391 return 0;
2392}
2393
2394int kvm_io_bus_insert_dev(struct kvm_io_bus *bus, struct kvm_io_device *dev,
2395 gpa_t addr, int len)
2396{
2397 if (bus->dev_count == NR_IOBUS_DEVS)
2398 return -ENOSPC;
2399
2400 bus->range[bus->dev_count++] = (struct kvm_io_range) {
2401 .addr = addr,
2402 .len = len,
2403 .dev = dev,
2404 };
2405
2406 sort(bus->range, bus->dev_count, sizeof(struct kvm_io_range),
2407 kvm_io_bus_sort_cmp, NULL);
2408
2409 return 0;
2410}
2411
2412int kvm_io_bus_get_first_dev(struct kvm_io_bus *bus,
2413 gpa_t addr, int len)
2414{
2415 struct kvm_io_range *range, key;
2416 int off;
2417
2418 key = (struct kvm_io_range) {
2419 .addr = addr,
2420 .len = len,
2421 };
2422
2423 range = bsearch(&key, bus->range, bus->dev_count,
2424 sizeof(struct kvm_io_range), kvm_io_bus_sort_cmp);
2425 if (range == NULL)
2426 return -ENOENT;
2427
2428 off = range - bus->range;
2429
2430 while (off > 0 && kvm_io_bus_sort_cmp(&key, &bus->range[off-1]) == 0)
2431 off--;
2432
2433 return off;
2434}
2435
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002436/* kvm_io_bus_write - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002437int kvm_io_bus_write(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002438 int len, const void *val)
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002439{
Sasha Levin743eeb02011-07-27 16:00:48 +03002440 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002441 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002442 struct kvm_io_range range;
2443
2444 range = (struct kvm_io_range) {
2445 .addr = addr,
2446 .len = len,
2447 };
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002448
2449 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002450 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2451 if (idx < 0)
2452 return -EOPNOTSUPP;
2453
2454 while (idx < bus->dev_count &&
2455 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2456 if (!kvm_iodevice_write(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002457 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002458 idx++;
2459 }
2460
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002461 return -EOPNOTSUPP;
2462}
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002463
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002464/* kvm_io_bus_read - called under kvm->slots_lock */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002465int kvm_io_bus_read(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2466 int len, void *val)
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002467{
Sasha Levin743eeb02011-07-27 16:00:48 +03002468 int idx;
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002469 struct kvm_io_bus *bus;
Sasha Levin743eeb02011-07-27 16:00:48 +03002470 struct kvm_io_range range;
2471
2472 range = (struct kvm_io_range) {
2473 .addr = addr,
2474 .len = len,
2475 };
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002476
Lai Jiangshan90d83dc2010-04-19 17:41:23 +08002477 bus = srcu_dereference(kvm->buses[bus_idx], &kvm->srcu);
Sasha Levin743eeb02011-07-27 16:00:48 +03002478 idx = kvm_io_bus_get_first_dev(bus, addr, len);
2479 if (idx < 0)
2480 return -EOPNOTSUPP;
2481
2482 while (idx < bus->dev_count &&
2483 kvm_io_bus_sort_cmp(&range, &bus->range[idx]) == 0) {
2484 if (!kvm_iodevice_read(bus->range[idx].dev, addr, len, val))
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002485 return 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002486 idx++;
2487 }
2488
Michael S. Tsirkinbda90202009-06-29 22:24:32 +03002489 return -EOPNOTSUPP;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002490}
2491
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002492/* Caller must hold slots_lock. */
Sasha Levin743eeb02011-07-27 16:00:48 +03002493int kvm_io_bus_register_dev(struct kvm *kvm, enum kvm_bus bus_idx, gpa_t addr,
2494 int len, struct kvm_io_device *dev)
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002495{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002496 struct kvm_io_bus *new_bus, *bus;
Gregory Haskins090b7af2009-07-07 17:08:44 -04002497
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002498 bus = kvm->buses[bus_idx];
Gregory Haskins090b7af2009-07-07 17:08:44 -04002499 if (bus->dev_count > NR_IOBUS_DEVS-1)
2500 return -ENOSPC;
2501
Thomas Meyer6da64fd2011-11-08 20:32:19 +01002502 new_bus = kmemdup(bus, sizeof(struct kvm_io_bus), GFP_KERNEL);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002503 if (!new_bus)
2504 return -ENOMEM;
Sasha Levin743eeb02011-07-27 16:00:48 +03002505 kvm_io_bus_insert_dev(new_bus, dev, addr, len);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002506 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2507 synchronize_srcu_expedited(&kvm->srcu);
2508 kfree(bus);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002509
2510 return 0;
2511}
2512
Marcelo Tosatti79fac952009-12-23 14:35:26 -02002513/* Caller must hold slots_lock. */
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002514int kvm_io_bus_unregister_dev(struct kvm *kvm, enum kvm_bus bus_idx,
2515 struct kvm_io_device *dev)
Gregory Haskins090b7af2009-07-07 17:08:44 -04002516{
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002517 int i, r;
2518 struct kvm_io_bus *new_bus, *bus;
Michael S. Tsirkin6c474692009-06-29 22:24:26 +03002519
Sasha Levincdfca7b2011-12-04 19:36:28 +02002520 bus = kvm->buses[bus_idx];
2521
2522 new_bus = kmemdup(bus, sizeof(*bus), GFP_KERNEL);
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002523 if (!new_bus)
2524 return -ENOMEM;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002525
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002526 r = -ENOENT;
2527 for (i = 0; i < new_bus->dev_count; i++)
Sasha Levin743eeb02011-07-27 16:00:48 +03002528 if (new_bus->range[i].dev == dev) {
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002529 r = 0;
Sasha Levin743eeb02011-07-27 16:00:48 +03002530 new_bus->dev_count--;
2531 new_bus->range[i] = new_bus->range[new_bus->dev_count];
2532 sort(new_bus->range, new_bus->dev_count,
2533 sizeof(struct kvm_io_range),
2534 kvm_io_bus_sort_cmp, NULL);
Gregory Haskins090b7af2009-07-07 17:08:44 -04002535 break;
2536 }
Marcelo Tosattie93f8a02009-12-23 14:35:24 -02002537
2538 if (r) {
2539 kfree(new_bus);
2540 return r;
2541 }
2542
2543 rcu_assign_pointer(kvm->buses[bus_idx], new_bus);
2544 synchronize_srcu_expedited(&kvm->srcu);
2545 kfree(bus);
2546 return r;
Gregory Haskins2eeb2e92007-05-31 14:08:53 -04002547}
2548
Avi Kivity774c47f2007-02-12 00:54:47 -08002549static struct notifier_block kvm_cpu_notifier = {
2550 .notifier_call = kvm_cpu_hotplug,
Avi Kivity774c47f2007-02-12 00:54:47 -08002551};
2552
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002553static int vm_stat_get(void *_offset, u64 *val)
Avi Kivityba1389b2007-11-18 16:24:12 +02002554{
2555 unsigned offset = (long)_offset;
Avi Kivityba1389b2007-11-18 16:24:12 +02002556 struct kvm *kvm;
2557
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002558 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002559 raw_spin_lock(&kvm_lock);
Avi Kivityba1389b2007-11-18 16:24:12 +02002560 list_for_each_entry(kvm, &vm_list, vm_list)
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002561 *val += *(u32 *)((void *)kvm + offset);
Jan Kiszkae935b832011-02-08 12:55:33 +01002562 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002563 return 0;
Avi Kivityba1389b2007-11-18 16:24:12 +02002564}
2565
2566DEFINE_SIMPLE_ATTRIBUTE(vm_stat_fops, vm_stat_get, NULL, "%llu\n");
2567
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002568static int vcpu_stat_get(void *_offset, u64 *val)
Avi Kivity1165f5f2007-04-19 17:27:43 +03002569{
2570 unsigned offset = (long)_offset;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002571 struct kvm *kvm;
2572 struct kvm_vcpu *vcpu;
2573 int i;
2574
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002575 *val = 0;
Jan Kiszkae935b832011-02-08 12:55:33 +01002576 raw_spin_lock(&kvm_lock);
Avi Kivity1165f5f2007-04-19 17:27:43 +03002577 list_for_each_entry(kvm, &vm_list, vm_list)
Gleb Natapov988a2ca2009-06-09 15:56:29 +03002578 kvm_for_each_vcpu(i, vcpu, kvm)
2579 *val += *(u32 *)((void *)vcpu + offset);
2580
Jan Kiszkae935b832011-02-08 12:55:33 +01002581 raw_spin_unlock(&kvm_lock);
Christoph Hellwig8b88b092008-02-08 04:20:26 -08002582 return 0;
Avi Kivity1165f5f2007-04-19 17:27:43 +03002583}
2584
Avi Kivityba1389b2007-11-18 16:24:12 +02002585DEFINE_SIMPLE_ATTRIBUTE(vcpu_stat_fops, vcpu_stat_get, NULL, "%llu\n");
2586
Alexey Dobriyan828c0952009-10-01 15:43:56 -07002587static const struct file_operations *stat_fops[] = {
Avi Kivityba1389b2007-11-18 16:24:12 +02002588 [KVM_STAT_VCPU] = &vcpu_stat_fops,
2589 [KVM_STAT_VM] = &vm_stat_fops,
2590};
Avi Kivity1165f5f2007-04-19 17:27:43 +03002591
Hamo4f69b682011-12-15 14:23:16 +08002592static int kvm_init_debug(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002593{
Hamo4f69b682011-12-15 14:23:16 +08002594 int r = -EFAULT;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002595 struct kvm_stats_debugfs_item *p;
2596
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002597 kvm_debugfs_dir = debugfs_create_dir("kvm", NULL);
Hamo4f69b682011-12-15 14:23:16 +08002598 if (kvm_debugfs_dir == NULL)
2599 goto out;
2600
2601 for (p = debugfs_entries; p->name; ++p) {
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002602 p->dentry = debugfs_create_file(p->name, 0444, kvm_debugfs_dir,
Avi Kivity1165f5f2007-04-19 17:27:43 +03002603 (void *)(long)p->offset,
Avi Kivityba1389b2007-11-18 16:24:12 +02002604 stat_fops[p->kind]);
Hamo4f69b682011-12-15 14:23:16 +08002605 if (p->dentry == NULL)
2606 goto out_dir;
2607 }
2608
2609 return 0;
2610
2611out_dir:
2612 debugfs_remove_recursive(kvm_debugfs_dir);
2613out:
2614 return r;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002615}
2616
2617static void kvm_exit_debug(void)
2618{
2619 struct kvm_stats_debugfs_item *p;
2620
2621 for (p = debugfs_entries; p->name; ++p)
2622 debugfs_remove(p->dentry);
Hollis Blanchard76f7c872008-04-15 16:05:42 -05002623 debugfs_remove(kvm_debugfs_dir);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002624}
2625
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002626static int kvm_suspend(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002627{
Alexander Graf10474ae2009-09-15 11:37:46 +02002628 if (kvm_usage_count)
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002629 hardware_disable_nolock(NULL);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002630 return 0;
2631}
2632
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002633static void kvm_resume(void)
Avi Kivity59ae6c62007-02-12 00:54:48 -08002634{
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002635 if (kvm_usage_count) {
Jan Kiszkae935b832011-02-08 12:55:33 +01002636 WARN_ON(raw_spin_is_locked(&kvm_lock));
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002637 hardware_enable_nolock(NULL);
Zachary Amsdenca84d1a2010-08-19 22:07:28 -10002638 }
Avi Kivity59ae6c62007-02-12 00:54:48 -08002639}
2640
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002641static struct syscore_ops kvm_syscore_ops = {
Avi Kivity59ae6c62007-02-12 00:54:48 -08002642 .suspend = kvm_suspend,
2643 .resume = kvm_resume,
2644};
2645
Izik Eiduscea7bb22007-10-17 19:17:48 +02002646struct page *bad_page;
Anthony Liguori35149e22008-04-02 14:46:56 -05002647pfn_t bad_pfn;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002648
Avi Kivity15ad7142007-07-11 18:17:21 +03002649static inline
2650struct kvm_vcpu *preempt_notifier_to_vcpu(struct preempt_notifier *pn)
2651{
2652 return container_of(pn, struct kvm_vcpu, preempt_notifier);
2653}
2654
2655static void kvm_sched_in(struct preempt_notifier *pn, int cpu)
2656{
2657 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2658
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002659 kvm_arch_vcpu_load(vcpu, cpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002660}
2661
2662static void kvm_sched_out(struct preempt_notifier *pn,
2663 struct task_struct *next)
2664{
2665 struct kvm_vcpu *vcpu = preempt_notifier_to_vcpu(pn);
2666
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002667 kvm_arch_vcpu_put(vcpu);
Avi Kivity15ad7142007-07-11 18:17:21 +03002668}
2669
Avi Kivity0ee75be2010-04-28 15:39:01 +03002670int kvm_init(void *opaque, unsigned vcpu_size, unsigned vcpu_align,
Rusty Russellc16f8622007-07-30 21:12:19 +10002671 struct module *module)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002672{
2673 int r;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002674 int cpu;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002675
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002676 r = kvm_arch_init(opaque);
2677 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002678 goto out_fail;
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002679
2680 bad_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2681
2682 if (bad_page == NULL) {
2683 r = -ENOMEM;
2684 goto out;
2685 }
2686
Anthony Liguori35149e22008-04-02 14:46:56 -05002687 bad_pfn = page_to_pfn(bad_page);
2688
Huang Yingbf998152010-05-31 14:28:19 +08002689 hwpoison_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2690
2691 if (hwpoison_page == NULL) {
2692 r = -ENOMEM;
2693 goto out_free_0;
2694 }
2695
2696 hwpoison_pfn = page_to_pfn(hwpoison_page);
2697
Gleb Natapovedba23e2010-07-07 20:16:45 +03002698 fault_page = alloc_page(GFP_KERNEL | __GFP_ZERO);
2699
2700 if (fault_page == NULL) {
2701 r = -ENOMEM;
2702 goto out_free_0;
2703 }
2704
2705 fault_pfn = page_to_pfn(fault_page);
2706
Avi Kivity8437a612009-06-06 14:52:35 -07002707 if (!zalloc_cpumask_var(&cpus_hardware_enabled, GFP_KERNEL)) {
Rusty Russell7f59f492008-12-07 21:25:45 +10302708 r = -ENOMEM;
2709 goto out_free_0;
2710 }
2711
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002712 r = kvm_arch_hardware_setup();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002713 if (r < 0)
Rusty Russell7f59f492008-12-07 21:25:45 +10302714 goto out_free_0a;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002715
Yang, Sheng002c7f72007-07-31 14:23:01 +03002716 for_each_online_cpu(cpu) {
2717 smp_call_function_single(cpu,
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002718 kvm_arch_check_processor_compat,
Jens Axboe8691e5a2008-06-06 11:18:06 +02002719 &r, 1);
Yang, Sheng002c7f72007-07-31 14:23:01 +03002720 if (r < 0)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002721 goto out_free_1;
Yang, Sheng002c7f72007-07-31 14:23:01 +03002722 }
2723
Avi Kivity774c47f2007-02-12 00:54:47 -08002724 r = register_cpu_notifier(&kvm_cpu_notifier);
2725 if (r)
Zhang Xiantaod23087842007-11-29 15:35:39 +08002726 goto out_free_2;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002727 register_reboot_notifier(&kvm_reboot_notifier);
2728
Rusty Russellc16f8622007-07-30 21:12:19 +10002729 /* A kmem cache lets us meet the alignment requirements of fx_save. */
Avi Kivity0ee75be2010-04-28 15:39:01 +03002730 if (!vcpu_align)
2731 vcpu_align = __alignof__(struct kvm_vcpu);
2732 kvm_vcpu_cache = kmem_cache_create("kvm_vcpu", vcpu_size, vcpu_align,
Joe Perches56919c52007-11-12 20:06:51 -08002733 0, NULL);
Rusty Russellc16f8622007-07-30 21:12:19 +10002734 if (!kvm_vcpu_cache) {
2735 r = -ENOMEM;
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002736 goto out_free_3;
Rusty Russellc16f8622007-07-30 21:12:19 +10002737 }
2738
Gleb Natapovaf585b92010-10-14 11:22:46 +02002739 r = kvm_async_pf_init();
2740 if (r)
2741 goto out_free;
2742
Avi Kivity6aa8b732006-12-10 02:21:36 -08002743 kvm_chardev_ops.owner = module;
Christian Borntraeger3d3aab12008-12-02 11:17:32 +01002744 kvm_vm_fops.owner = module;
2745 kvm_vcpu_fops.owner = module;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002746
2747 r = misc_register(&kvm_dev);
2748 if (r) {
Mike Dayd77c26f2007-10-08 09:02:08 -04002749 printk(KERN_ERR "kvm: misc device register failed\n");
Gleb Natapovaf585b92010-10-14 11:22:46 +02002750 goto out_unreg;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002751 }
2752
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002753 register_syscore_ops(&kvm_syscore_ops);
2754
Avi Kivity15ad7142007-07-11 18:17:21 +03002755 kvm_preempt_ops.sched_in = kvm_sched_in;
2756 kvm_preempt_ops.sched_out = kvm_sched_out;
2757
Hamo4f69b682011-12-15 14:23:16 +08002758 r = kvm_init_debug();
2759 if (r) {
2760 printk(KERN_ERR "kvm: create debugfs files failed\n");
2761 goto out_undebugfs;
2762 }
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002763
Avi Kivityc7addb92007-09-16 18:58:32 +02002764 return 0;
Avi Kivity6aa8b732006-12-10 02:21:36 -08002765
Hamo4f69b682011-12-15 14:23:16 +08002766out_undebugfs:
2767 unregister_syscore_ops(&kvm_syscore_ops);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002768out_unreg:
2769 kvm_async_pf_deinit();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002770out_free:
Rusty Russellc16f8622007-07-30 21:12:19 +10002771 kmem_cache_destroy(kvm_vcpu_cache);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002772out_free_3:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002773 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity774c47f2007-02-12 00:54:47 -08002774 unregister_cpu_notifier(&kvm_cpu_notifier);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002775out_free_2:
Zhang Xiantaod23087842007-11-29 15:35:39 +08002776out_free_1:
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002777 kvm_arch_hardware_unsetup();
Rusty Russell7f59f492008-12-07 21:25:45 +10302778out_free_0a:
2779 free_cpumask_var(cpus_hardware_enabled);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002780out_free_0:
Gleb Natapovedba23e2010-07-07 20:16:45 +03002781 if (fault_page)
2782 __free_page(fault_page);
Huang Yingbf998152010-05-31 14:28:19 +08002783 if (hwpoison_page)
2784 __free_page(hwpoison_page);
Zhang Xiantaod23087842007-11-29 15:35:39 +08002785 __free_page(bad_page);
Avi Kivityca45aaa2007-03-01 19:21:03 +02002786out:
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002787 kvm_arch_exit();
Zhang Xiantaod23087842007-11-29 15:35:39 +08002788out_fail:
Avi Kivity6aa8b732006-12-10 02:21:36 -08002789 return r;
2790}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002791EXPORT_SYMBOL_GPL(kvm_init);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002792
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002793void kvm_exit(void)
Avi Kivity6aa8b732006-12-10 02:21:36 -08002794{
Darrick J. Wong0ea4ed82009-10-14 16:21:00 -07002795 kvm_exit_debug();
Avi Kivity6aa8b732006-12-10 02:21:36 -08002796 misc_deregister(&kvm_dev);
Rusty Russellc16f8622007-07-30 21:12:19 +10002797 kmem_cache_destroy(kvm_vcpu_cache);
Gleb Natapovaf585b92010-10-14 11:22:46 +02002798 kvm_async_pf_deinit();
Rafael J. Wysockifb3600c2011-03-23 22:16:23 +01002799 unregister_syscore_ops(&kvm_syscore_ops);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002800 unregister_reboot_notifier(&kvm_reboot_notifier);
Avi Kivity59ae6c62007-02-12 00:54:48 -08002801 unregister_cpu_notifier(&kvm_cpu_notifier);
Takuya Yoshikawa75b71272010-11-16 17:37:41 +09002802 on_each_cpu(hardware_disable_nolock, NULL, 1);
Zhang Xiantaoe9b11c12007-11-14 20:38:21 +08002803 kvm_arch_hardware_unsetup();
Zhang Xiantaof8c16bb2007-11-14 20:40:21 +08002804 kvm_arch_exit();
Rusty Russell7f59f492008-12-07 21:25:45 +10302805 free_cpumask_var(cpus_hardware_enabled);
Huang Yingbf998152010-05-31 14:28:19 +08002806 __free_page(hwpoison_page);
Izik Eiduscea7bb22007-10-17 19:17:48 +02002807 __free_page(bad_page);
Avi Kivity6aa8b732006-12-10 02:21:36 -08002808}
Zhang Xiantaocb498ea2007-11-14 20:39:31 +08002809EXPORT_SYMBOL_GPL(kvm_exit);